Dementia methylation and end-motif profiles in cell-free DNA and uses thereof

Abstract

The present application, in certain aspects, relates to methods of determining a dementia methylation profile and / or a dementia status based on methylation information, including information of one or more non-CpG methylation loci and / or methylation sequence context, from cell-free DNA (cfDNA) from a blood-based sample from an individual. The present application, in certain other aspects, relates to methods of determining a dementia status based on end-motif sequence information from cfDNA from a blood-based sample from an individual. In certain other aspects, the application is also directed to uses of a dementia methylation profile or dementia status, such as in methods of diagnosis and / or treatment, computer systems, and non-transitory computer-readable storage media.

Description

Attorney Docket No.: 347372000140DEMENTIA METHYLATION AND END-MOTIF PROFILES IN CELL-FREE DNA AND USES THEREOFCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Application No. 63 / 712,931, filed on October 28, 2024, the contents of which are hereby incorporated herein by reference in their entirety.TECHNICAL FIELD

[0002] The present application, in certain aspects, relates to methods of determining a dementia methylation profile and / or a dementia status based on methylation information, including information of one or more non-CpG methylation loci, taking into account methylation sequence context, from cell-free DNA (cfDNA) from a blood-based sample from an individual. The present application, in certain other aspects, relates to methods of determining a dementia status based on end-motif sequence information from cfDNA from a blood-based sample from an individual. In certain other aspects, the application is also directed to uses of a dementia methylation profile and / or dementia status, such as in methods of diagnosis and / or treatment, computer systems, and non-transitory computer-readable storage media.BACKGROUND

[0003] Dementia, a health condition that affects memory and the function of the brain, is a leading cause of death globally and a significant cause of morbidity. Alzheimer’s disease and Parkinson’s disease are the most common types of dementia. Early detection and molecular characterization of dementia is crucial for implementing treatments sooner, such as in a pre-symptomatic or early-stage patients, and for developing novel drugs for treatment. However, clinical detection of dementia often occurs at a late stage as investigations to diagnose dementia commence following clinical suspicion of significant cognitive decline. Definitive signs and symptoms can be non-specific, and are often missed by clinicians, in the early stages. Despite this, molecular studies have demonstrated that brain pathology can occur years before symptoms. Dementia research has been general limited by the challenge of1MF-363846971Attorney Docket No.: 347372000140 sampling brain tissue directly. Routine brain biopsies are not obtained to detect or study neurodegenerative disease pathology. Furthermore, biopsies obtained are confounded by the biopsy process itself, as their viability may be limited. As a result, available brain tissue studies are biased towards post-mortem sampling, which offer insight into advanced disease but cannot effectively generalize to the biology of early-stage disease. Blood-based protein assays for dementia have emerged recently, enabling non-invasive detection. However, in their current form, protein-only assays provide only a single disease metric without molecular characterization on a genomic, epigenomic, or transcriptomic level. Furthermore, the specificity of such approaches can be variable since they are largely based on the presence of amyloid, which can be found in healthy individuals. Thus, there remains a need in the field for techniques to study dementia using non-invasively acquired sample and that are useful for providing insights into early-stage disease.SUMMARY OF THE INVENTION

[0004] In some aspects, provided herein is a method of determining a dementia methylation profile from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determining the dementia methylation profile based on the methylation information. In some embodiments, the method further comprises comparing the dementia methylation profile to a reference profile. In some embodiments, the method further comprises determining a dementia status based on the comparison of the dementia methylation profile to the reference profile. In some embodiments, the dementia status indicates that the individual has a dementia.

[0005] In other aspects, provided herein is a method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting, using the one or more processors, methylation information from the methylation-processed sequence read2MF-363846971Attorney Docket No.: 347372000140 information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output the dementia status.

[0006] In some embodiments, the one or more non-CpG methylation loci comprises one or more CHH methylation loci and / or one or more CHG methylation loci, where H is any of A, T, or C. In some embodiments, each of the one or more CHH methylation loci is 5'-CAA- 3', 5'-CAT-3', 5'-CAC-3', 5'-CTA-3', 5'-CTT-3', 5'-CTC-3', 5'-CCA-3', 5'-CCT-3', or 5'-CCC- 3'. In some embodiments, each of the one or more CHG methylation loci is 5'-CAG-3', 5'- CTG-3', or 5'-CCG-3'. In some embodiments, the one or more methylation loci comprises 5'- CAC-3'.

[0007] In some embodiments, the methylation information comprises the methylation state of each or a subset of the one or more non-CpG methylation loci from the methylation- processed sequence read information mapping to the one or more DMRs. In some embodiments, the methylation information comprises a percent methylated for each or the subset of the one or more non-CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs. In some embodiments, the methylation information further comprises information from the methylation-processed sequence read information of one or more CpG methylation loci mapping to the one or more DMRs. In some embodiments, the methylation information comprises the methylation state of each of the one or more CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs. In some embodiments, the methylation information comprises a percent methylated for each of the one or more CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs.

[0008] In some embodiments, the methylation information is compiled in a text format. In some embodiments, the methylation information is compiled into a two-dimensional data structure. In some embodiments, the two-dimensional data structure is a dataframe.

[0009] In some embodiments, the methylation information further comprises sequence context. In some embodiments, the sequence context is one or more bases on the 5' and / or 3' side a methylation locus, e.g., 5'-ACAC-3', 5'-CCAC-3', or 5'-TCAC-3'.3MF-363846971Attorney Docket No.: 347372000140

[0010] In some embodiments, the methylation information comprises information of a plurality of non-CpG locus types in the one or more DMRs. In some embodiments, the methylation information comprises information for one non-CpG locus type in the one or more DMRs. In some embodiments, the methylation information comprises information for a plurality of sequence context types in the one or more DMRs. In some embodiments, the methylation information comprises information for one sequence context type in the one or more DMRs. In some embodiments, the method further comprises obtaining the one or more DMRs. In some embodiments, the method further comprises determining the one or more DMRs. In some embodiments, the determining the one or more DMRs is based on one or more of CpG methylation loci, nil context, one or more non-CpG methylation loci, sequence context, one or more non-CpG methylation loci with sequence context, a tissue type, a cell type, or a disease. In some embodiments, the tissue type is a brain tissue. In some embodiments, the brain tissue is the hippocampus, frontal lobe, basal ganglia, entorhinal cortex, cingulate cortex or temporal lobe. In some embodiments, the brain tissue is the frontal lobe, parietal lobe, temporal lobe, occipital lobe, insula, hippocampus, amygdala, thalamus, hypothalamus, epithalamus, subthalamus, basal ganglia, caudate nucleus, putamen, globus pallidus, nucleus accumbens, substantia nigra, red nucleus, superior colliculus, inferior colliculus, cerebellum, pons, medulla oblongata, midbrain, corpus callosum, cingulate cortex, prefrontal cortex, motor cortex, somatosensory cortex, visual cortex, auditory cortex, brainstem, pituitary gland, or pineal gland. In some embodiments, the cell type is a brain cell. In some embodiments, the brain cell is a neuron, oligodendrocyte, astrocyte, pericyte, neural stem cell, or microglia. In some embodiments, the disease is Alzheimer’s disease or Parkinson’s disease. In some embodiments, the disease is Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, amyotrophic lateral sclerosis (ALS), frontotemporal dementia (FTD), Lewy body dementia, vascular dementia, corticobasal degeneration, progressive supranuclear palsy, multiple system atrophy, spinocerebellar ataxia, Friedreich’s ataxia, prion disease (including Creutzfeldt-Jakob disease), spinal muscular atrophy, Batten disease, Niemann-Pick disease, metachromatic leukodystrophy, Wilson’s disease, Alzheimer’s disease with cerebral amyloid angiopathy, or posterior cortical atrophy.

[0011] In other aspects, provided herein is a method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, sequence read information from a4MF-363846971Attorney Docket No.: 347372000140 sequencing analysis of the cfDNA; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status. In some embodiments, the 5'- and / or 3' terminal end motifs comprise at least 2 bases in length. In some embodiments, the 5'- and / or 3' terminal end motifs are 3 bases in length. In some embodiments, the 5'- and / or 3' terminal end motifs are 4 bases in length. In some embodiments, the information regarding 5'- and / or 3' terminal end motifs is the frequency of each unique end motif.

[0012] In some embodiments, the trained machine learning model comprises a classifier. In some embodiments, the trained machine learning model comprises a tree-based classifier. In some embodiments, the trained machine learning model is an ensemble classifier. In some embodiments, the trained machine learning model comprises a gradient boosting algorithm. In some embodiments, the trained machine learning model comprises a CatBoost algorithm. In some embodiments, the trained machine learning model comprises a Random Forest classifier. In some embodiments, the method further comprises training a machine learning model to obtain the trained machine learning model.

[0013] In some embodiments, the method further comprises obtaining a blood, plasma, cerebral spinal fluid (CSF), or serum sample from the individual.

[0014] In some embodiments, the method further comprises performing the sequencing analysis on the cfDNA. In some embodiments, the sequencing analysis is performed using a methylation-processed sequencing analysis. In some embodiments, the sequencing analysis is an enzyme-based methylation sequencing technique. In some embodiments, the enzymebased methylation sequencing technique is EM-SEQ™. In some embodiments, the sequencing analysis is performed to a depth of about 5x to about lOOOx. In some embodiments, the sequencing analysis is a single-end mode sequencing technique or a paired- end mode sequencing technique. In some embodiments, the sequencing analysis is a wholegenome sequencing analysis. In some embodiments, the sequencing analysis is a targeted sequencing analysis. In some embodiments, following receiving the sequence read information, the method further comprises one or more steps of quality control or sequence5MF-363846971Attorney Docket No.: 347372000140 trimming. In some embodiments, following receiving the sequence read information, the method further comprises mapping to a reference genome.

[0015] In some embodiments, the dementia status indicates the individual has the dementia. In some embodiments, the dementia is Alzheimer’s disease or Parkinson’s disease. In some embodiments, the dementia is of a pre-symptomatic stage, early stage, moderate stage, or advanced stage.

[0016] In other aspects, provided herein is a method of diagnosing a dementia in an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting methylation information from the sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); determining a dementia methylation profile based on the methylation information; and comparing the dementia methylation profile to a reference profile to determining a dementia status thereby diagnosing the dementia in the individual.

[0017] In other aspects, provided herein is a method of diagnosing a dementia in an individual, the method comprising: receiving, at one or more processors, methylation- processed sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from the individual; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non- CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status thereby diagnosing the dementia in the individual.

[0018] In other aspects, provided herein is a method of diagnosing a dementia in an individual, the method comprising: receiving, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status6MF-363846971Attorney Docket No.: 347372000140 thereby diagnosing the dementia in the individual. In some embodiments, the dementia status indicates that the individual has the dementia.

[0019] In other aspects, provided herein is a method of treating an individual having a dementia, the method comprising: diagnosing the individual as having the dementia according to any method described herein; and administering to the individual a therapeutic agent for the dementia. In some embodiments, the therapeutic agent is selected from the group consisting of a cholinesterase inhibitor, glutamate regulator, N-methyl-D-aspartate (NMD A) antagonist, acetylcholinesterase inhibitor, antidepressant, antipsychotic, PU.l inhibitory modulator, and anticonvulsant. In some embodiments, the therapeutic agent is selected from the group consisting of donepezil, galantamine, rivastigmine, lecanemab, memantine, brexpiprazole, aducanumab, donanemab, lecanemab, posiphen, RASRxl902, RASRxl911, any of glycosylated-angiotensin 1-7, MW071, MW109, EHI-16, M3, LM11A- 31, CT1812, BPN14770, allopregnanolone, MW151, and MW150. In some embodiments, the dementia is Alzheimer’s disease or Parkinson’s disease. In some embodiments, the dementia is of a pre-symptomatic stage, early stage, moderate stage, or advanced stage.

[0020] In other aspects, provided herein is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: obtaining methylation- processed sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from an individual; extracting methylation information from the methylation- processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determining a dementia methylation profile based on the methylation information. In some embodiments, one or more programs further include instructions for comparing the dementia methylation profile to a reference profile. In some embodiments, the one or more programs further include instructions for determining a dementia status based on the comparison of the dementia methylation profile to the reference profile.

[0021] In other aspects, provided herein is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more7MF-363846971Attorney Docket No.: 347372000140 processors, the one or more programs including instructions for: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status.

[0022] In other aspects, provided herein is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: receiving, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from an individual; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output a dementia status.

[0023] In other aspects, provided herein is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: obtain methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determine a dementia methylation profile based on the methylation information. In some embodiments, the one or more programs comprising instructions, which when executed by the one or more processors of the system, further cause the system to compare the dementia methylation profile to a reference profile. In some embodiments, the one or more programs comprising instructions, which when executed by the one or more processors of the system, further cause the system to determine a dementia status based on the comparison of the dementia methylation profile to the reference profile.8MF-363846971Attorney Docket No.: 347372000140

[0024] In other aspects, provided herein is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: receive, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and input, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status.

[0025] In other aspects, provided herein is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: receive, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and input, using the one or more processors, the end-motif sequence information into a trained machine learning model to output a dementia status.BRIEF DESCRIPTION OF THE DRAWINGS

[0026] FIG. 1 shows a high-level schematic of an example workflow of a methodology taught herein involving methylation information of one or more non-CpG methylation loci and, optionally, methylation sequence context in one or more DMRs.

[0027] FIG. 2 shows a high-level schematic of an example workflow of a methodology taught herein involving end-motif sequence information.

[0028] FIG. 3A shows a heatmap displaying the methylation level across multiple tissues. White in the heatmap represent a higher proportion of methylation levels at a given differentially methylated region (DMR).

[0029] FIG. 3B shows a bar graph displaying the number of significant DMRs for each methylation region types that was assayed.9MF-363846971Attorney Docket No.: 347372000140

[0030] FIG. 4A shows boxplots displaying the classification performance of CpG and non-CpG methylation, including with certain methylation sequence contexts, across DMRs using neuronal regions as the area under curve (AUC) of receiver operating characteristic (ROC) curves. FIGS. 4B and 4C show ROC curves for CpG and non-CpG methylation sequence contexts. FIG. 4D shows ROC curves for end motif frequency per sample used as input into a machine learning classifier, which generated an AUC for classification of 0.78.DETAILED DESCRIPTION OF THE DISCLOSURE

[0031] Provided herein, in certain aspects, are methods of determining a dementia methylation profile and / or a dementia status based on (i) methylation information, including information of one or more non-CpG methylation loci and / or methylation sequence context, or (ii) end-motif sequence information. The disclosure of this application is based, at least in part, on the inventors’ findings and unique perspectives regarding methylation and end-motif- based approaches for analyzing cell-free DNA (cfDNA) from blood-based samples, e.g., plasma, for the classification of a dementia, such as Alzheimer’s disease, in an individual. Prior to the inventions taught herein, sample-based detection and / or classification of dementia in an individual was based on difficult-to-obtain brain samples (which are often only available post-mortem) or non-invasive approaches that suffered from a lack of much needed dimensionality as such techniques are based on proteins and thus only provide a single disease metric. Moreover, the specificity of such prior approaches was low because tracked proteins (e.g., amyloid) can also be found in individuals not suffering from a dementia. As taught herein, the inventors discovered how to process and use non-CpG methylation, sequence context, and end-motif sequence information to obtain a higher-degree of specificity and sensitivity for detecting and / or classifying a dementia in an individual. The methods taught herein leverage low-abundance signals in cell-free DNA from a blood-based sample indicative of certain states of dementia, including early-stage development, with methodologies designed to significantly enhancing signal-to-noise.

[0032] Thus, in certain aspects, provided herein is a method of determining a dementia methylation profile from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting methylation information10MF-363846971Attorney Docket No.: 347372000140 from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determining the dementia methylation profile based on the methylation information. In some embodiments, the method further comprises comparing the dementia methylation profile to a reference profile. In some embodiments, the method further comprises determining a dementia status based on the comparison of the dementia methylation profile to the reference profile. In some embodiments, the dementia status indicates that the individual has a dementia.

[0033] In other aspects, provided herein is a method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output the dementia status.

[0034] In other aspects, provided herein is a method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, sequence read information from a sequencing analysis of the cfDNA; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status.

[0035] In other aspects, provided herein is a method of diagnosing a dementia in an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting methylation information from the sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); determining a dementia methylation profile based on the methylation information; and11MF-363846971Attorney Docket No.: 347372000140 comparing the dementia methylation profile to a reference profile to determining a dementia status thereby diagnosing the dementia in the individual.

[0036] In other aspects, provided herein is a method of diagnosing a dementia in an individual, the method comprising: receiving, at one or more processors, methylation- processed sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from the individual; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non- CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status thereby diagnosing the dementia in the individual.

[0037] In other aspects, provided herein is a method of diagnosing a dementia in an individual, the method comprising: receiving, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status thereby diagnosing the dementia in the individual.

[0038] In other aspects, provided herein is a method of treating an individual having a dementia, the method comprising: diagnosing the individual as having the dementia according to any method described herein; and administering to the individual a therapeutic agent for the dementia.

[0039] In other aspects, provided herein is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: obtaining methylation- processed sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from an individual; extracting methylation information from the methylation- processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determining a dementia methylation12MF-363846971Attorney Docket No.: 347372000140 profile based on the methylation information. In some embodiments, one or more programs further include instructions for comparing the dementia methylation profile to a reference profile. In some embodiments, the one or more programs further include instructions for determining a dementia status based on the comparison of the dementia methylation profile to the reference profile.

[0040] In other aspects, provided herein is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status.

[0041] In other aspects, provided herein is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: receiving, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from an individual; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output a dementia status.

[0042] In other aspects, provided herein is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: obtain methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci13MF-363846971Attorney Docket No.: 347372000140 in one or more differentially methylated regions (DMRs); and determine a dementia methylation profile based on the methylation information. In some embodiments, the one or more programs comprising instructions, which when executed by the one or more processors of the system, further cause the system to compare the dementia methylation profile to a reference profile. In some embodiments, the one or more programs comprising instructions, which when executed by the one or more processors of the system, further cause the system to determine a dementia status based on the comparison of the dementia methylation profile to the reference profile.

[0043] In other aspects, provided herein is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: receive, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and input, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status.

[0044] In other aspects, provided herein is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: receive, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and input, using the one or more processors, the end-motif sequence information into a trained machine learning model to output a dementia status.

[0045] All publications, including patent documents, scientific articles and databases, referred to in this application are hereby incorporated herein by reference in their entirety for all purposes to the same extent as if each individual publication were individually incorporated by reference.14MF-363846971Attorney Docket No.: 347372000140

[0046] The section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.I. Definitions

[0047] Unless defined otherwise, all terms of art, notations and other technical and scientific terms or terminology used herein are intended to have the same meaning as is commonly understood by one of ordinary skill in the art to which the claimed subject matter pertains. In some cases, terms with commonly understood meanings are defined herein for clarity and / or for ready reference, and the inclusion of such definitions herein should not necessarily be construed to represent a substantial difference over what is generally understood in the art. If a definition set forth herein is contrary to or otherwise inconsistent with a definition set forth in the patents, applications, published applications and other publications that are incorporated herein by reference, the definition set forth herein prevails over the definition that is incorporated herein by reference.

[0048] As used herein, “treatment,” treating,” or “treat,” or equivalents thereof, are defined as an approach for obtaining beneficial or desired results including clinical results. For purposes of this application, beneficial or desired clinical results include, but are not limited to, one or more of the following: alleviating one or more symptoms resulting from a dementia, such as Alzheimer’s disease or Parkinson’s disease, diminishing the extent of a disorder, stabilizing a disorder (e.g., preventing or delaying the worsening of a disorder), delay or slowing the progression of a disorder, ameliorating a disorder state, decreasing the dose of one or more other medications required to treat a disorder, delaying the progression of a disorder, increasing or improving the quality of life, and / or prolonging survival. Also encompassed by “treatment” is a reduction of pathological consequence of a disorder. The methods herein contemplate any one or more of these aspects of treatment.

[0049] In some embodiments, the term “diagnosis,” or equivalents thereof, are used herein to refer to the identification (including presence or absence) and / or classification of a molecular or pathological state, disease, or condition. For example, diagnosis may refer to identification of a particular type of a dementia, e.g., Alzheimer’s disease or Parkinson’s disease. Diagnosis may also refer to the classification of a particular type of a dementia, such as by staging or impacts on certain aspects of the individual.15MF-363846971Attorney Docket No.: 347372000140

[0050] As used herein, a “subject” or an “individual,” which are terms that are used interchangeably, is a mammal. In some embodiments, a “mammal” includes humans, nonhuman primates, domesticated and farm animals, and zoo, sports, or pet animals, such as dogs, horses, rabbits, cattle, pigs, hamsters, gerbils, mice, ferrets, rats, cats, and monkeys. In some embodiments, the individual is human.

[0051] As used herein, the terms “comprising” (and any form or variant of comprising, such as “comprise” and “comprises”), “having” (and any form or variant of having, such as “have” and “has”), “including” (and any form or variant of including, such as “includes” and “include”), or “containing” (and any form or variant of containing, such as “contains” and “contain”), are inclusive or open-ended and do not exclude additional, un-recited additives, components, integers, elements, or method steps.

[0052] Throughout this disclosure, various aspects of the claimed subject matter are presented in a range format. It should be understood that the description in range format is merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the claimed subject matter. Accordingly, the description of a range should be considered to have specifically disclosed all the possible sub-ranges as well as individual numerical values within that range. For instance, where a range of values is provided, it is understood that each intervening value, to the tenth of the unit of the lower limit, unless the context clearly dictate otherwise, between the upper and lower limit of that range and any other stated or intervening value in that stated range, is encompassed within the disclosure, subject to any specifically excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the disclosure. In some embodiments, two opposing and open-ended ranges are provided for a feature, and in such description it is envisioned that combinations of those two ranges are provided herein. For example, in some embodiments, it is described that a feature is greater than about 10 units, and it is described (such as in another sentence) that the feature is less than about 20 units, and thus, the range of about 10 units to about 20 units is described herein.

[0053] The term “about” as used herein refers to the usual error range for the respective value readily known in this technical field. Reference to “about” a value or parameter herein includes (and describes) variations that are directed to that value or parameter per se. For16MF-363846971Attorney Docket No.: 347372000140 example, description referring to “about X” includes description of “X.” Exemplary degrees of error are within 20 percent (%), such as within 15%, within 10%, or within 5% of a given value or range of values.

[0054] As used herein, including in the appended claims, the singular forms “a,” “or,” and “the” include plural referents unless the context clearly dictates otherwise. For example, “a” or “an” means “at least one” or “one or more.” It is understood that aspects and variations described herein include embodiments “consisting of’ and / or “consisting essentially of’ such aspects and variations.

[0055] The term “and / or” as used herein a phrase such as “A and / or B” is intended to include both A and B; A or B; A (alone); and B (alone). Likewise, the term “and / or” as used herein a phrase such as “A, B, and / or C” is intended to encompass each of the following embodiments: A, B, and C; A, B, or C; A or C; A or B; B or C; A and C; A and B; B and C; A (alone); B (alone); and C (alone).

[0056] Those skilled in the art will recognize that several embodiments are possible within the scope and spirit of the present disclosure. The following description illustrates the disclosure and, of course, should not be construed in any way as limiting the scope of the inventions described herein.II. Methods

[0057] Provided herein, in certain aspects, are methods of determining a dementia methylation profile and / or a dementia status based on (i) methylation information, including information of one or more non-CpG methylation loci and, optionally, sequence context, or (ii) end-motif sequence information. FIGS. 1 and 2 are provided to illustrate high-level example methodologies provided herein. Such illustrations are for purposes of teaching certain aspects of inventions disclosed herein and are not intended to limit the scope of the claimed methodologies. For example, embodiments provided herein may include different orderings, fewer or additional steps, and / or different steps. As shown in the workflow 100 of FIG. 1, in some embodiments, the methods provided herein comprise obtaining methylation- processed sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from an individual 102. As described herein, a diverse array of sample preparation and sequencing techniques are known to be suitable for obtaining such17MF-363846971Attorney Docket No.: 347372000140 information from cfDNA from a blood-based sample (such as from techniques involving chemical or enzymatic modification of non-methylated cytosines). In the workflow 100 of FIG. 1, using the methylation-processed sequence read information, the methods provided herein may comprise extracting methylation information, e.g., based on one or more non- CpG methylation loci and, optionally, methylation sequence context, in one or more differentially methylated regions (DMRs) 104. The extracted methylation information can be formatted in a suitable manner for subsequent use and may contain all or a subset of the methylation information contained in the methylation-processed sequence read information. In some embodiments, the extracted methylation information contains a metric describing one or more methylation sites, such as methylation state or methylated percentage. As shown in FIG. 1, the methylation information is then used to determine a dementia methylation profile and / or a dementia status (e.g., indicating that an individual has a dementia) 106. In some embodiments, such determination is conducted with the use of a machine learning model.

[0058] As shown in the workflow 200 of FIG. 2, in some embodiments, the methods provided herein comprise obtaining sequence read information (e.g., methylation-processed sequence read information) from a sequencing analysis of cfDNA from a blood-based sample of an individual 202. As described herein, a diverse array of sample preparation and sequencing techniques are known to be suitable for obtaining such information from cfDNA from a blood-based sample, including techniques involving, or not, chemical or enzymatic modification of non-methylated cytosines. In the workflow 200 of FIG. 2, using the sequence read information the methods provided herein may comprise extracting end-motif sequence information of reads in the sequence read information 104. As described herein, end motifs may be characterized by a certain number of bases at the 5'- and / or 3' end of a read, such as 3 or 4 bases. The end-motif sequence information can be formatted in a suitable manner for subsequent use and may contain all or a subset of the end-motif sequence information contained in the sequence read information. In FIG. 2, the end-motif information is then used to determine a dementia status (e.g., indicating that an individual has a dementia) 206. In some embodiments, such determination is conducted with the use of a machine learning model.

[0059] Thus, provided herein, in certain aspects, is a method of determining a dementia methylation profile from cell-free DNA (cfDNA) from a blood-based sample obtained from18MF-363846971Attorney Docket No.: 347372000140 an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of the cfDNA sample; extracting methylation information from the methylation-processed sequence read information for one or more non- CpG methylation loci in one or more differentially methylated regions (DMRs); and determining the dementia methylation profile based on the methylation information. In some embodiments, the method further comprises comparing the dementia methylation profile to a reference profile or a threshold. In some embodiments, the method further comprises determining a dementia status based on the comparison of the dementia methylation profile to the reference profile or threshold, e.g., wherein the determining is based on probabilistic methodologies. In some embodiments, the dementia status indicates that the individual has a dementia.

[0060] In some aspects, provided herein is a method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output the dementia status.

[0061] In some aspects, provided herein is a method of determining a dementia methylation profile from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting methylation information from the methylation-processed sequence read information for one or more methylation sequence contexts, e.g., including of a CpG methylation locus and / or a non-CpG methylation locus, in one or more differentially methylated regions (DMRs); and determining the dementia methylation profile based on the methylation information. In some embodiments, the method further comprises comparing the dementia methylation profile to a reference profile. In some embodiments, the method comprises determining a dementia status based on the comparison of the dementia methylation profile to the reference profile. In some embodiments, the dementia status indicates that the individual has a dementia.19MF-363846971Attorney Docket No.: 347372000140

[0062] In some aspects, provided herein is a method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more methylation sequence contexts, e.g., including of a CpG locus and / or a non-CpG locus, in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output the dementia status.

[0063] In some aspects, provided herein is a method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, sequence read information from a sequencing analysis of the cfDNA; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status.

[0064] In some embodiments, the information used in the methods provided herein, such as non-CpG methylation loci, CpG methylation loci, sequence context, and end motifs, are based on a desired AUC, which may be based on a clinical context, e.g., the classification of a specific dementia. The term “AUC” as used herein is an abbreviation for the “area under a curve”. In particular it refers to the area under a Receiver Operating Characteristic (ROC) curve. ’The ROC curve is a plot of the true positive rate against the false positive rate for the different possible cut points of a methodology. It shows the trade-off between sensitivity and specificity depending on the selected cut point (any increase in sensitivity will be accompanied by a decrease in specificity). The area under an ROC curve (AUC) is a measure for the accuracy of methodology described herein (the larger the area the better; the optimum is 1; a random test would have a ROC curve lying on the diagonal with an area of 0.5). Eg, J . P. Egan. (1975) Signal Detection Theory and ROC Analysis, Academic Press, New York, which is hereby incorporated herein by reference in its entirety.20MF-363846971Attorney Docket No.: 347372000140

[0065] For purposes of brevity, aspects of the methods disclosed herein are described in more detail below in a modular fashion. Such presentation is not to be construed as limiting the scope of combinations of the various aspects encompassed by the disclosure of the present application to form a method as taught herein.A. Aspects of methylation and sequence context

[0066] DNA methylation in eukaryotes typically occurs via the attachment of a methyl group at the C5-position of the nucleotide base cytosine (C). In some embodiments, methylation of cytosine occurs at CpG dinucleotides loci (5'-CG-3'), where G is guanine. Cytosine methylation can also occur at non-CpG methylation loci such as 5'-CHG-3' and 5'- CHH-3', where H is adenine (A), cytosine, or thymine (T). Examples of non-CpG methylation loci include 5'-CAA-3', 5'-CAT-3', 5'-CAC-3', 5'-CTA-3', 5'-CTT-3', 5'-CTC-3', 5'-CCA-3', 5'-CCT-3', 5'-CCC-3', 5'-CAG-3', 5'-CTG-3', and 5'-CCG-3'. Sequence context, as used herein, refers to base pairs on the 5' and / or 3' side of a methylation locus, e.g., one or two bases on the 5' and / or 3' side of a methylation locus. As taught in the methods provided herein, methylation information, including information from non-CpG methylation loci, and, optionally, sequence context, can be used to obtain a dementia methylation profile and / or a dementia status (e.g., indicating the individual has, or a probability thereof, or does not have a dementia) Additional details for both methylation information and sequence context, and how such information is obtained and used in the methods provided herein are further discussed in, e.g., the sections below. i. Methylation information

[0067] In certain aspects provided herein, the methods involve methylation information as obtained from a sequencing analysis of cfDNA from a blood-based sample from an individual. In some embodiments, the methylation information is obtained from a methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs). In some embodiments, the one or more non-CpG methylation loci comprises one or more CHH methylation loci and / or one or more CHG methylation loci, where H is any of A, T, or C. In some embodiments, each of the one or more CHH methylation loci is 5'-CAA-3', 5'-CAT-3', 5'-CAC-3', 5'-CTA-3', 5'-CTT-3',21MF-363846971Attorney Docket No.: 3473720001405'-CTC-3', 5'-CCA-3', 5'-CCT-3', or 5'-CCC-3'. In some embodiments, each of the one or more CHG methylation loci is 5'-CAG-3', 5'-CTG-3', or 5'-CCG-3'.

[0068] In some embodiments, the methylation information comprises the methylation state (the presence or absence of a methylation) of each or a subset of the one or more non- CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs. In some embodiments, the methylation information comprises a percent methylated (e.g., as determined based on the number of methylated occurrences of a locus over the total reads covering said locus) for each or the subset of the one or more non- CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs.

[0069] In some embodiments, the methylation information further comprises information from the methylation-processed sequence read information of one or more CpG methylation loci mapping to the one or more DMRs. In some embodiments, the methylation information comprises the methylation state (the presence or absence of a methylation) of each of the one or more CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs. In some embodiments, the methylation information comprises a percent methylated (e.g., as determined based on the number of methylated occurrences of a locus over the total reads covering said locus) for each of the one or more CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs.

[0070] In some embodiments, the sequence read information may contain information regarding a diverse array of methylation loci, including different types of loci. In some embodiments, the methylation information may contain all of said information or a subset of said information. In some embodiments, the design of downstream analyses will guide the information that is included in the methylation information. In some embodiments, the methylation information comprises information of a plurality of non-CpG loci types in the one or more DMRs. In some embodiments, the methylation information comprises information for one non-CpG locus type in the one or more DMRs. In some embodiments, the sequence read information contained in methylation information has been filtered, such as based on read quality, read coverage, or the number of methylation loci on a read. For22MF-363846971Attorney Docket No.: 347372000140 example, in some embodiments, the methods provided herein comprise filtering raw sequence read information for reads having 2 or more, such as 3 or more, methylation loci.

[0071] In some embodiments, the methylation information further comprises information regarding sequence context. As described herein, sequence context refers to base pairs on the 5' and / or 3' side of a methylation locus (the limits of which are defined by, e.g., CG, CHH, or CHG). In some embodiments, the sequence context is one or more bases on the 5' and / or 3' side a methylation locus. In some embodiments, the sequence context is one to five bases on the 5' and / or 3' side a methylation locus. In some embodiments, the sequence context is one or two bases on the 5' and / or 3' side a methylation locus.

[0072] In some embodiments, sequence context may be established on a per methylation locus or per methylation locus type basis. For example, in some embodiments, the sequence context for a CpG methylation locus is two bases, which can be positioned in any configuration, e.g., one base on the 5' and one base on the 3' side. In some embodiments, the sequence context for a non-CpG methylation locus is one base, such as one base on the 5' or one base on the 3' side. In some embodiments, the sequence context, as described in the number of bases relative to the methylation locus, may be the same for all methylation loci. In some embodiments, the sequence context, as described in the number of bases relative to the methylation locus, is different for two methylation loci (such methylation loci may be defined positionally, e.g., in a different position in the genome, or by type, e.g., CpG versus non-CpG loci). In some embodiments, the sequence context is defined by sequence context type ( / .< ., a methylation locus type having the same context, e.g., AC AC is one sequence context type). In some embodiments, the methylation information comprises information for a plurality of sequence context types in the one or more DMRs. In some embodiments, the methylation information comprises information for one sequence context type in the one or more DMRs. In some embodiments, the methylation information comprises the methylation state (the presence or absence of a methylation) of each or a subset of the total sequence contexts from the methylation-processed sequence read information mapping to the one or more DMRs. In some embodiments, the methylation information comprises a percent methylated (e.g., as determined based on the number of methylated occurrences of a sequence context type over the total reads covering said sequence context type) for each or the subset of the total sequence contexts from the methylation-processed sequence read information mapping to the one or more DMRs.23MF-363846971Attorney Docket No.: 347372000140

[0073] In some embodiments, the methylation information comprises information of one or more methylation loci without sequence context and information of one or more methylation loci with sequence context. In some embodiments, the methylation information comprises information of one or more methylation loci without sequence context. In some embodiments, the methylation information comprises information of one or more methylation loci with sequence context. In some embodiments, the methods provided herein comprise selecting one or more methylation loci and, optionally, one or more sequence contexts.

[0074] In some embodiments, the methylation information used in the methods provided herein comprises information of one or more of 5'-ACAC-3', 5'-CAC-3', 5'-CCAC-3', 5'- GCAC-3', or 5'-TCAC-3'. In some embodiments, the methylation information used in the methods provided herein comprises information of one or more of 5'-ACAC-3'. In some embodiments, the methylation information used in the methods provided herein further comprises information of one or more of 5'-CGG-3', 5'-CGC-3', 5'-CGT-3', or 5'-CGA-3'.

[0075] The form of methylation information may take many forms and may be guided by downstream uses of said information, e.g., use in a machine learning model. In some embodiments, the methylation information is compiled in a text format, e.g., a JSON file. In some embodiments, the methylation information is compiled into a two-dimensional data structure, e.g., a table. In some embodiments, the two-dimensional data structure is a dataframe. In some embodiments, the methylation information is compiled as one or more of a table, matrix, dictionary, or array.Hi. Differentially methylation regions (DMRs)

[0076] The methods provided herein, in certain aspects, comprises the determination and / or use of differentially methylated regions (DMRs). In some embodiments, the methods provided herein comprise selecting one or more DMRs. As used herein, DMRs is used in reference to a subset or subsets of the genome where the DNA methylation pattern may differ between certain groups (e.g., disease versus healthy). DMRs often identify regions with critical importance to gene regulation. Therefore, DMRs can be critical to the understanding of disease mechanisms, developmental biology, cell state, cell regulation, or cell function.

[0077] In some embodiments, the method further comprises obtaining the one or more DMRs, e.g., based on known DMRs or previously identified DMRs. In some embodiments,24MF-363846971Attorney Docket No.: 347372000140 the method further comprises determining the one or more DMRs, such as based on a characteristic desired to be evaluated. In some embodiments, the determining the one or more DMRs is based on one or more of CpG methylation loci, nil context, one or more non-CpG methylation loci, sequence context, one or more non-CpG methylation loci with sequence context, a tissue type, a cell type, or a disease. In some embodiments, the tissue type is a brain tissue. In some embodiments, the brain tissue is from the hippocampus, parahippocampal regions, putamen, thalamus, cerebral peduncle, cingulum, striatum, motor cortex, white matter, corpus collosum, internal capsule, cingulum, posterior cingulate cortex, temporal lobe, temporal pole, frontal lobe, inferior frontal gyrus medial frontal gyrus, and / or occipital lobe. In some embodiments, the brain tissue is the frontal lobe, parietal lobe, temporal lobe, occipital lobe, insula, hippocampus, amygdala, thalamus, hypothalamus, epithalamus, subthalamus, basal ganglia, caudate nucleus, putamen, globus pallidus, nucleus accumbens, substantia nigra, red nucleus, superior colliculus, inferior colliculus, cerebellum, pons, medulla oblongata, midbrain, corpus callosum, cingulate cortex, prefrontal cortex, motor cortex, somatosensory cortex, visual cortex, auditory cortex, brainstem, pituitary gland, or pineal gland. In some embodiments, the cell type is a brain cell. In some embodiments, the brain cell is a neuron, excitatory neuron, inhibitory neuron, dopaminergic neuron, glial cell, astrocyte, oligodendrocyte, oligodendrocyte precursor cell, microglia, immune cell, or epithelial cell. In some embodiments, the disease is a dementia, such as disclosed herein, e.g., Alzheimer’s disease or Parkinson’s disease. In some embodiments, the disease is Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, amyotrophic lateral sclerosis (ALS), frontotemporal dementia (FTD), Lewy body dementia, vascular dementia, corticobasal degeneration, progressive supranuclear palsy, multiple system atrophy, spinocerebellar ataxia, Friedreich’s ataxia, prion disease (including Creutzfeldt-Jakob disease), spinal muscular atrophy, Batten disease, Niemann-Pick disease, metachromatic leukodystrophy, Wilson’s disease, Alzheimer’s disease with cerebral amyloid angiopathy, or posterior cortical atrophy. Methods of obtaining DMRs are known in the art, e.g. Loyfer, et al., biorXiv. 2024. In some embodiments, DMRs are manually identified by dividing the genome into 100b, Ikb, lOkb, lOOkb, 1Mb, 10Mb, or 100Mb windows. In some embodiments, the DMRs are used to filter sequence reads based on attributes of the reads mapping thereto. In some embodiments, the DMR filtering comprises a minimum number of methylation loci per read threshold. In some embodiments, the minimum number of methylation loci per read threshold is 1, 2, 3, 4,25MF-363846971Attorney Docket No.: 347372000140 or 5. In some embodiments, the minimum number of methylation loci per read threshold is 3. In some embodiments, the DMR filtering comprises filtering out reads containing a SNP. In some embodiments, the DMR filtering comprises filtering out reads to repetitive regions (in some embodiments referred to as repeat masking). In some embodiments, the selection of DMRs is based on a minimum methylation difference between target and background, such as between a disease state and a non-disease state, e.g., healthy, respectively. In some embodiments, the DMRs are selected based on information from a tissue sample, In such embodiments, the minimum methylation difference between target and background may be 0.3 or more, such as 0.4 or more, 0.5 or more, or 0.6 or more. In some embodiments, the minimum methylation difference between target and background is 0.5. In some embodiments, the DMRs are selected based on information from a blood-based sample comprising cfDNA, such as plasma or serum. In such embodiments, the cfDNA from brain tissue may be a low percentage of the total cfDNA, and thus a lower difference between target and background may be used, such as about 0.001 or more, including about any of 0.01 or more, 0.05 or more, or 0.1 or more.B. End-motif profiles

[0078] Provided herein, in certain aspects, are methods of determining a dementia status (e.g., indicating that an individual has a dementia) based on end-motif sequence information. End motifs are sets of bases (e.g., two or more) at the 5'- and / or 3' terminal ends of a sequence read, and as demonstrated herein said unique sets can be used to classify a dementia in an individual.

[0079] In some embodiments, provided herein is a method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, sequence read information from a sequencing analysis of the cfDNA; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status.26MF-363846971Attorney Docket No.: 347372000140

[0080] In some embodiments, the 5'- and / or 3' terminal end motifs comprise at least 2 bases in length, such as any of 2 bases, 3 bases, 4 bases, 5 bases, 6 bases, 7 bases, 8 bases, 9 bases, or 10 bases. In some embodiments, the 5'- and / or 3' terminal end motifs are 3 bases in length. In some embodiments, the 5'- and / or 3' terminal end motifs are 4 bases in length. In some embodiments, the 5'-terminal end motifs are 3 bases in length. In some embodiments, the 5'-terminal end motifs are 4 bases in length. In some embodiments, the 3'-terminal end motifs are 3 bases in length. In some embodiments, the 3 '-terminal end motifs are 4 bases in length. In some embodiments, the 5'- and 3'-terminal end motifs are 3 bases in length. In some embodiments, the 5'- and 3'-terminal end motifs are 4 bases in length.

[0081] In some embodiments, the information regarding 5'- and / or 3' terminal end motifs is the frequency of each unique end motif. The form of end motif information may take many forms and may be guided by downstream uses of said information, e.g., use in a machine learning model. In some embodiments, the end motif information is compiled in a text format, e.g., a JSON file. In some embodiments, the end motif information is compiled into a two-dimensional data structure, e.g., a table. In some embodiments, the two-dimensional data structure is a dataframe. In some embodiments, the end motif information is compiled as one or more of a table, matrix, dictionary, or array.C. Machine learning models

[0082] The methods provided herein, in certain aspects, comprises machine learning models. As used herein, machine learning model is used in reference to a group of computational methods in which a system learns from data to perform a specific task without explicit instructions of the entire process. A machine learning model can be trained on a dataset to perform a specific task before introducing a novel dataset for the machine learning model to perform on. Machine learning models encompass many distinct methods with the capability of receiving complex data, such as medical data, and informing on complex states, such as a disease or condition.

[0083] The machine learning models disclosed herein may be used to predict a dementia status. More specifically, the models may be configured to receive methylation information and / or end-motif sequence information and output a score (e.g., a probability or a class label) that is indicative of dementia status. The machine learning models used herein employ a27MF-363846971Attorney Docket No.: 347372000140 combination of techniques to reduce complexity, provide more efficient and accurate training, and achieve improved memory and computing resource usage.

[0084] To predict a dementia status, methylation information and / or end-motif sequence information is provided to a machine learning model. Optionally, the methylation information and / or end-motif sequence information is normalized or scaled by coverage, read-depth, and / or context abundance before providing the information to the model, which can improve model calibration. The model to which the received information is provided can be a classifier and can be tree-based and ensemble-compatible in order to handle heterogeneous tabular features and deliver robust feature importance. For example, the model can be a Random Forest model, a Gradient Boosting Model, or a CatBoost model, among others. The output of the model is a dementia status score, such as a probability or a class label. Thresholds associated with the score can be set via ROC / AUC analysis. The score can be reported as a presence or absence of dementia and can optionally be aligned to stage categories (e.g., pre-symptomatic, early, moderate, or advanced).

[0085] In some embodiments, the machine learning model is selected from the group consisting of supervised learning models, unsupervised learning models, reinforcement learning models, semi-supervised learning models, self-supervised learning models, multiinstance learning models, inductive learning models, deductive learning models, transductive learning models, multi-task learning models, active learning models, online learning models, transfer learning models, ensemble learning models, or deep learning models. In some embodiments, the supervised learning model is a classification model. In some embodiments, the classification model is selected from a group of a deterministic model, a decision tree model, a probabilistic model, a hidden Markov model (HMM), a mixture model, a Gaussian mixture model, a Bayesian network model (e.g., a naive Bayes classifier), an artificial neural network model, a support vector machine (SVM) model, or a K-nearest neighbor model. In some embodiments the supervised learning model is a regression model. In some embodiments the regression model is selected from the group consisting of a linear model, a logistic model, a polynomial model, a stepwise model, a decision tree model, a random forest model, a SVM model, a ridge model, a lasso model, an ElasticNet model, or a Bayesian linear model. In some embodiments the unsupervised learning model is selected from an exclusive clustering model, an overlapping clustering model, a hierarchical clustering model, a probabilistic clustering model, an association model, or a dimensionality reduction model.28MF-363846971Attorney Docket No.: 347372000140In some embodiments, the machine learning model further comprises an ensemble model. In some embodiments, the ensemble model comprises one or more of a random forest, a boosting, a gradient boosting, a light gradient boosting, a CatBoost, a bagging, a stacking, or an XGBoosting model. In some embodiments, the machine learning model comprises a deep learning model (e.g., an artificial neural network (ANN), a convolutional neural network (CNN), or a recurrent neural network (RNN)). In some embodiments, the machine learning model comprises a transformer. Such types of machine learning models are known in the art, e.g. Ahsan et al., Healthcare (Basel), 2022. In some embodiments, the machine learning model receives a dataset in which different features are associated with a different weight. In some embodiments, the machine learning model assigns different weights to different features of the input data. i. Ensemble learning models

[0086] In some embodiments, the machine learning model further comprises an ensemble learning model. Ensemble learning refers to the process in which multiple machine learning models are combined to a single machine learning model to improve performance. Ensemble learning models are designed to overcome the specificity and sensitivity of any single model alone. In some embodiments, the ensemble learning model comprises one or more of a random forest, a boosting, a gradient boosting, a CatBoost, a bagging, a stacking, or a XGBoosting.

[0087] In some embodiments, the ensemble learning model comprises bagging. Bagging, or bootstrap aggregating, refers to an ensemble learning approach in which a dataset is bootstrapped before using each dataset subset to train a separate machine learning model. Bootstrapping refers to the process in which a dataset is split into multiple datasets by random sampling with replacement. Once the individual machine learning models are trained on their respective datasets, their predictions can be averaged or otherwise combined in order to produce an output. Using a bagging approach is advantageous because it reduces model variance, thus improving accuracy and generalizability. In some embodiments, the bagging is a random forest. A random forest refers to a bagging approach in which multiple decision trees are combined into a single model. Where a typical bagging approach simply bootstraps the dataset, a random forest approach further divides features of the dataset into random29MF-363846971Attorney Docket No.: 347372000140 subsets. The goal of dividing the features into random subsets is to encourage the decision trees that comprise a random forest to be as independent and diverse as possible.

[0088] In some embodiments, the ensemble learning model comprises boosting. Boosting refers to an ensemble learning approach in which multiple models are trained sequentially. Each model is then intended to improve on the errors of the previous model. In some embodiments, the boosting is a gradient boosting. Gradient boosting refers to a boosting approach in which a gradient descent is used to minimize a loss function. Gradient descent refers to an optimization algorithm in which model parameters are iteratively changed towards the steepest descent of a loss function. In some embodiments, the gradient boosting is a CatBoost. CatBoost refers to a gradient boosting library that is specifically designed for categorical features. ii. Training

[0089] The machine learning models described herein may be trained or the methods provided herein may encompass training. A training for a machine learning model refers to the process in which a model is taught to make predictions, classifications, or decisions based on an input data. Training involves the use of a training dataset, a validation dataset, and testing dataset. In some embodiments, a dataset is split into a training and validation dataset and a testing dataset. In some embodiments, a dataset is split in a 80 to 20 ratio where 80% of the dataset is used for training and validation and 20% of the dataset is used for testing.

[0090] The generalizability of the machine learning model during training can be improved through the use of cross-validation. Cross-validation is a model evaluation technique that divides a dataset into multiple subsets, known as “folds.” The method systematically rotates which folds are used for training and which are used for validation. For example, in some embodiments, the training comprises a k-fold cross-validation. In a k-fold cross-validation, a training dataset is evenly split into k equal-sized subsets where k is an integer. The training is then repeated k number of times in which one subset is left out for validation and the remaining is combined for training. In some embodiments, the k-fold cross-validations comprises at least 2-folds, 3-folds, 4-folds, 5-folds, 10-folds, 15-folds, 20- folds, or 25-folds. In some embodiments, the k-fold cross-validations comprises 5-folds. In some embodiments, final training and validation metrics are obtained by averaging all repeats30MF-363846971Attorney Docket No.: 347372000140 of the k-fold cross-validation. In some embodiments, the final training and validation metrics are used to determine the parameters of the machine learning model. Using cross-validation techniques is advantageous because it improves model reliability. Instead of relying on one random split of the dataset, cross-validation tests the model across multiple different folds of data. This provides greater insight into how the model will perform. Additionally, training and testing the model on multiple folds derived from the same dataset makes efficient use of limited data and helps prevent overfitting or underfitting of the model.

[0091] According to some aspects, the machine learning models described herein may be trained using adaptive optimizers. Adaptive optimizers are algorithms that can automatically adjust the learning rate for each model parameter during training of a machine learning model. The learning rates may be adjusted based on gradients observed during training. In some examples, the adaptive optimizers may include AdaGrad (Adaptive Gradient Algorithm), RMSProp (Root Mean Square Propagation), Adam (Adaptive Moment Estimation), or variants or combinations thereof. Automatically adjusting learning rates during training can result in faster and more stable training.

[0092] Alternatively or additionally, learning rates may be adjusted during training using learning rate schedulers. Learning rate schedulers are algorithms that automatically adjust a model’s learning rate during training based on a predefined schedule or based on training progress. While adaptive optimizers adjust the learning rate for each parameter during training, learning rate schedulers adjust the learning rate for the entire model during training. In some examples, a learning rate scheduler may reduce the learning rate of a model as training progresses in order to ensure that the model converges quickly in the beginning but can be finely adjusted as the model stabilizes.

[0093] According to some aspects, the model training may utilize mixed precision methods. Mixed precision methods use multiple numerical formats (e.g., 16-bit and 32-bit floating point) to represent weights, activations, and gradients during training. This reduces the amount of memory and computing resources required for training. For example, a traditional training process may use 32-bit floating point values for the entire training process, while a mixed precision method may identify the steps that require full precision and use 32-bit floating point values for only those steps and use 16-bit floating point values for31MF-363846971Attorney Docket No.: 347372000140 the steps that do not require full precision. This is advantageous because 16-bit values require less memory and can be processed more quickly than 32-bit values.

[0094] Quantization techniques may be used to reduce the numerical precision of certain parameters. For example, quantization may be used to map 32-bit floating point values to 16- bit floating point values. The lower-precision values may require less memory and less processing power, which can improve model efficiency and reduce storage requirements.

[0095] According to some aspects, training the machine learning models disclosed herein may include pruning the machine learning model during or after training. Pruning refers to the systematic removal of unnecessary parts (e.g., parameters, layers, etc.) of a machine learning model. Pruning reduces a model’s size and memory requirements, thereby improving the model’s efficiency while maintaining its accuracy. For example, a machine learning model may be trained, and the trained model may be evaluated to identify relatively unimportant parameters (e.g., via a sensitivity analysis). The identified parameters may be removed, and the model may be re-trained or fine-tuned to recovery any lost accuracy.

[0096] According to some aspects, the machine learning models disclosed herein may be trained using training data represented as sparse matrices. Sparse matrices store only nonzero matrix entries and their positions, which conserves memory and computational resources required for processing the data. Using sparse matrices during model training improves the computational efficiency of the training process, which can increase the model’s learning rate and enable training at scale.

[0097] In some examples, the training data used to train the machine learning models disclosed herein may be imbalanced. For example, a training dataset may include multiple “classes” of data (e.g., data points representing positive diagnoses and data points representing negative diagnoses), and one class may be overrepresented relative to others. If unaddressed, training the model on this dataset can lead to a bias toward the dominant class(es). To ensure that the classes are represented more equally during training, balanced sampling techniques may be employed. In some examples, balanced sampling includes randomly removing samples from a majority class so that the number of samples in the majority class more closely matches the number of samples in the minority class(es). In other examples, balanced sampling includes duplicating or generating samples pertaining to the minority class(es). This can reduce model bias.32MF-363846971Attorney Docket No.: 347372000140

[0098] In other examples, imbalanced training datasets can be addressed using data augmentation. Data augmentation refers to increasing the size of a dataset by applying various transformations to existing samples in the dataset to generate new samples. Generating new samples improves the generalizability of the model trained using the enhanced dataset and improves model performance on unseen data.

[0099] The machine learning models disclosed herein may optionally be trained using distributed training techniques. Distributed training is the process of training a machine learning model across multiple different computing resources. Distributing the training process across multiple processors can improve training speed and efficiency, since individual processors may have limited memory and computing power that leads to longer processing times.

[0100] Training time and complexity may be reduced by using transfer learning techniques. Transfer learning refers to adapting an existing model that has been pre-trained on one task to be used to perform a new, different task. For example, parameters of the pretrained model that capture general patterns can be kept fixed, while other parameters can be modified and fine-tuned using data specific to the new task. Using transfer learning techniques can reduce the amount of data, time, and computing resources required to train a model as compared to training the model from scratch.D. Samples and sequencing

[0101] The methods provided herein are based on sequence information obtained from sequencing cfDNA from a blood-based sample from an individual. In certain aspects, the description herein is contemplated that the term cfDNA encompass derivatives thereof, such as would occur due to, e.g., a sample processing step or a sequencing preparation step.

[0102] In some embodiments, the methods provided herein comprise obtaining a bloodbased sample from an individual. In some embodiments, the blood-based sample is a whole blood sample. In some embodiments, the blood-based sample is a plasma sample comprising cfDNA. In some embodiments, the blood-based sample is a serum sample comprising cfDNA. In some embodiments, the methods provided herein comprise obtaining a plasma sample from an individual. In some embodiments, the methods provided herein comprise obtaining a serum sample from an individual. Techniques for obtaining blood, plasma, or33MF-363846971Attorney Docket No.: 347372000140 serum samples and handling of said samples to preserve biological components thereof are known in the art, e.g., Tuck et al., J Proteome Res, 8, 2009 and Danesi et al., Clinica Chimica Acta, 520, 2021, the disclosures of each of which are hereby incorporated herein by reference in its entirety. In some embodiments, the method comprises generating a plasma or a serum sample comprising cfDNA from a blood-based sample from the individual.

[0103] In some embodiments, the blood-based sample, e.g., plasma, obtained from the individual is about 100 pL to about 5 mL, including any of 250 pL, 500 pL, 750 pL, 1 mL, 1.5 mL, 2 mL, 2.5 mL, 3 mL, 3.5 mL, 4 mL, 4.5 mL, or 5 mL. In some embodiments, the method comprises obtaining one or more blood and / or plasma and / or serum samples from an individual, which can be obtain roughly at the same time, e.g., such as would occur in single visit to the clinician, and / or over a time course.

[0104] In some embodiments, the methods provided herein comprise obtaining, such as generating, a plasma sample comprising cfDNA. In some embodiments, the methods provided herein comprise extracting nucleic acids from a blood-based sample. In some embodiments, the method comprises subjecting a blood-based sample from an individual to a nucleic acid enrichment step such as a to obtain cfDNA therein. Techniques and products for obtaining cfDNA from a blood-based sample are known in the art, e.g., the Qiagen Circulating Nucleic Acid kit, QIAAMP. In some embodiments, the method comprises admixing proteinase K to the blood-based sample, or a derivative thereof. In some embodiments, the methods comprise quantifying the amount of cfDNA.

[0105] In some embodiments, the sample is a cerebral spinal fluid (CSF) sample. Techniques are known for collecting CSF, including a lumbar puncture. In some embodiments, the obtained CSF sample is about 1 mL to about 10 mL, such as about any of 2 mL, 2.5 mL, 3 mL, 3.5 mL, 4 mL, 4.5 mL, 5 mL, 5.5 mL, 6 mL, 6.5 mL, 7 mL, 7.5 mL, 8 mL, 8.5 mL, 9 mL, or 9.5 mL.

[0106] In some embodiments, the methods provided herein are based on a methylation- processed process. Processing to preserve methylation information upon sequencing is well known in the art, and include techniques such as bisulfite sequencing, enzymatic approaches, e.g., EM-SEQ™, and methylation sensitive probe-based methodologies. In some embodiments, bisulfite sequence comprises treating cfDNA with a bisulfite reagent (e.g., sodium bisulfite) to produce bisulfite converted cfDNA (the bisulfite reagent converts34MF-363846971Attorney Docket No.: 347372000140 unmethylated cytosines to uracils while methylated cytosines remain unaffected). Subsequently, bisulfite-compatible primers may be used to amplify the cfDNA which converts uracils to thymidines, thereby allowing of the detection of methylated and unmethylated bases upon sequencing. In some embodiments, the enzymatic approach, e.g., EM-SEQ™ comprises enzymatically modifying methylated cytosines to prevent deamination in further enzymatic steps and then enzymatically converting unmethylated cytosines to uracils. Similar to bisulfite sequencing, PCR amplification then converts uracils to thymines prior to sequencing. In some embodiments, enzymatically modifying methylated cytosines is performed using TET2 and / or T4-BGT. In some embodiments, the non-disruptive methylation sequencing technique comprises enzymatically converting unmethylated cytosines to uracil using AP0BEC3 A. E.g., Vaisvila et aL, Genome Res, 31, 2021, which is hereby incorporated herein by reference in its entirety. In some embodiments, the sequencing technique that preserves methylation information is selected from the group consisting of whole-genome bisulfite sequencing (WGBS), reduced representation bisulfite sequencing (RRBS), oxidative bisulfite sequencing (oxBS-seq), TET-assisted bisulfite sequencing (TAB- seq), enzymatic methyl-seq (EM-seq), TAPS (TET-assisted pyridine borane sequencing), post-bi sulfite adaptor tagging (PBAT), targeted bisulfite sequencing, methyl-C capture sequencing (MethylCap-seq), methylated DNA immunoprecipitation sequencing (MeDIP- seq), methyl-CpG binding domain sequencing (MBD-seq), single-cell bisulfite sequencing (scBS-seq), nanopore native methylation sequencing (Oxford Nanopore Technologies, ONT), PacBio single-molecule real-time (SMRT) methylation detection, nanoEM or nanoMe-seq (native enzymatic methylation detection using nanopore), and enzymatic methylation calling from direct DNA current signals.

[0107] In some embodiments, the methods provided herein comprise performing library preparation, such as performing next generation library preparation. In some embodiments, the next generation sequencing library preparation technique comprises a step of end preparation. In some embodiments, the next generation sequencing library preparation technique comprises a step of adaptor ligation. In some embodiments, the next generation sequencing library preparation technique comprises a step of cleaning up adaptor ligated DNA. In some embodiments, library preparation may occur prior to a methylation processing step. In some embodiments, the method comprises adding a control to the sample, e.g, prior35MF-363846971Attorney Docket No.: 347372000140 to performing any enzymatic conversion steps. In some embodiments, the methods provided herein comprise quantifying the formed libraries, such as via qPCR.

[0108] In some embodiments, multiple samples, such as from one individual or a plurality of individual, may be multiplexed. In such embodiments, the multiplexed samples may be pooled prior to sequencing.

[0109] In some embodiments, the methods provided herein comprise sequencing cfDNA from a blood-based sample. In some embodiments, the sequencing analysis is a wholegenome sequencing analysis. In some embodiments, the sequencing analysis is a targeted sequencing analysis. In some embodiments, the sequencing is a single-end mode sequencing technique or a paired-end mode sequencing technique. Sequencing techniques, including next generation sequencing techniques, and instruments useful therefor are well known in the art. For example, in some embodiments, the sequencing is performed by an Illumina platform, such as NOVASEQ®. Suitable next generation sequencing technologies are widely available. Examples include the 454 Life Sciences platform (Roche, Branford, CT) (Margulies et a!.. Nature, 437, 2005); Illumina’s Genome Analyzer; QX200™ Droplet Digital™ PCR System from Bio-Rad; or DNA Sequencing by Ligation, SOLiD System (Applied Biosystems / Life Technologies; U.S. Pat. Nos. 6,797,470, 7,083,917, 7,166,434, 7,320,865, 7,332,285, 7,364,858, and 7,429,453 (Barany et al); the Helicos True Single Molecule DNA sequencing technology (Harris et al, 2008 Science, 320, 106-109; U.S. Pat. Nos. 7,037,687 and 7,645,596 (Williams et al); 7, 169,560 (Lapidus et al); 7,769,400 (Harris)), the single molecule, realtime (SMRT™) technology of Pacific Biosciences, and sequencing (Soni and Meller, 2007, Clin. Chem.53, 1996-2001); semiconductor sequencing (Ion Torrent; Personal Genome Machine); DNA nanoball sequencing; sequencing using technology from Dover Systems (Polonator), and technologies that do not require amplification or otherwise transform native DNA prior to sequencing (e.g., Pacific Biosciences and Helicos), such as nanopore-based strategies (e.g., Oxford Nanopore, Genia Technologies, and Nabsys). In some embodiments, the sequencing is a short-read whole-genome sequencing, a long-read whole-genome sequencing, or a hybrid whole-genome sequencing.

[0110] In some embodiments, the sequence read information is obtained from a sequencing reaction performed to a depth, such as a mean depth, of about lx to about lOOOx, such as any of about 5x to about 50x, about 5x to about 25x, or about lOx to about 20x. In36MF-363846971Attorney Docket No.: 347372000140 some embodiments, the sequencing depth is about any of 5x, lOx, 15x, 20x, 25x, 30x, 35x, 40x, 45x, 50x, 60x, 70x, 80x, 90x, lOOx, 125x, 150x, 175x, 200x, 225x, 250x, 275x, 300x, 325x, 350x, 375x, 400x, 425x, 450x, 475x, 500x, 550x, 600x, 650x, 700x, 750x, 800x, 850x, 900x, 950x, or lOOOx.[OHl] In some embodiments, the methods provided herein comprise one or more steps of filtering, quality control, and / or sequence trimming. For example, in some embodiments, following receiving the methylation-processed sequence read information, the method further comprises one or more steps of quality control or sequence trimming. In some embodiments, following receiving the methylation-processed sequence read information, the method further comprises mapping to a reference genome. In some embodiments, such steps are performed prior to extraction of information, such as methylation information.E. Dementia

[0112] The methods provided herein, in certain aspects, find use in assessing an individual suspected of having a dementia or an individual known to have a dementia. As use herein, a dementia is used in reference to a group of diseases affecting one or more parts of the brain wherein an individual experiences cognitive decline. Dementia is often characterized by the significant decline of at least one of the cognition domains, including executive function, complex attention, language, learning, memory, perceptual-motor, or social cognition.

[0113] In some embodiments, the dementia is selected from the group consisting of Alzheimer disease, vascular dementia, frontotemporal lobar degeneration dementia, Lewy body dementia, Parkinson’s disease, HIV infection-based dementia, Huntington disease, prion disease-based dementia, substance and or medication use-based dementia, traumatic brain injury-based dementia, another medical condition-based dementia, multiple etiologies- based dementia, and unspecified dementia. Such types of dementia, as well as symptoms associate therewith, are known in the art, e.g., Emmady et al., StatPearls, 2022, which is hereby incorporated herein by reference in its entirety. In some embodiments, the dementia is of a pre-symptomatic stage, early stage, moderate stage, or advanced stage. In some embodiments, the individual exhibits one or more symptoms associated with a dementia. In some embodiments, the individual has been diagnosed with a dementia based on a technique37MF-363846971Attorney Docket No.: 347372000140 other than taught herein, such as based on a review of medical and cognitive history and / or an assessment such as the Montreal Cognitive Assessment, e.g., Arvanitakis et al., JAMA, y , 2019.F. Methods of diagnosis and treatment

[0114] In certain aspects, the methods provided herein are used in a method of diagnosis or a method of treatment. In some embodiments, the methods are based on determining the presence or absence of a dementia. In some embodiments, the methods are based on determining a likelihood of the presence or absence of a dementia. In some embodiments, the methods are based on determining the severity or staging of a dementia. In some embodiments, the methods involve monitoring an aspect of a dementia or the treatment thereof.

[0115] In some embodiments, provided is a method of diagnosing a dementia (such as determining the presence or absence thereof) in an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting methylation information from the sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); determining a dementia methylation profile based on the methylation information; and comparing the dementia methylation profile to a reference profile to determining a dementia status thereby diagnosing the dementia in the individual.

[0116] In some embodiments, provided is a method of diagnosing a dementia (such as determining the presence or absence thereof) in an individual, the method comprising: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status thereby diagnosing the dementia in the individual.38MF-363846971Attorney Docket No.: 347372000140

[0117] In some embodiments, provided is a method of diagnosing a dementia (such as determining the presence or absence thereof) in an individual, the method comprising: receiving, at one or more processors, sequence read information from a sequencing analysis of a cfDNA from a blood-based sample from the individual; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status thereby diagnosing the dementia in the individual.

[0118] In some embodiments, the dementia status indicates that the individual has the dementia.

[0119] In some embodiments, provided is a method of treating an individual having a dementia, the method comprising: diagnosing the individual as having the dementia according to any one of the methods provided herein; and administering to the individual a therapeutic agent for the dementia. In some embodiments, the therapeutic agent is selected from the group consisting of a cholinesterase inhibitor, glutamate regulator, N-methyl-D- aspartate (NMD A) antagonist, acetylcholinesterase inhibitor, antidepressant, antipsychotic, PU.l inhibitory modulator, and anticonvulsant. In some embodiments, the therapeutic agent is selected from the group consisting of donepezil, galantamine, rivastigmine, lecanemab, memantine, brexpiprazole, aducanumab, donanemab, lecanemab, posiphen, RASRxl902, RASRxl911, any of glycosylated-angiotensin 1-7, MW071, MW109, EHI-16, M3, LM11A- 31, CT1812, BPN14770, allopregnanolone, MW151, and MW150. In some embodiments, the therapeutic agent is a monoclonal antibody, an immunotherapy, a cell therapy and / or a cellular rejuvenation therapy, or a cellular replacement therapy, including neuronal and microglial replacement and / or modification.III. Computer Systems and component thereof

[0120] Further provided herein are computer systems and aspects useful therefor such as networks (e.g., non-transitory computer-readable storage mediums) designed to implement any of the methods taught herein.39MF-363846971Attorney Docket No.: 347372000140

[0121] In some aspects, provided is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: obtaining methylation- processed sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from an individual; extracting methylation information from the methylation- processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determining a dementia methylation profile based on the methylation information. In some embodiments, the one or more programs further include instructions for comparing the dementia methylation profile to a reference profile. In some embodiments, the one or more programs further include instructions for determining a dementia status based on the comparison of the dementia methylation profile to the reference profile.

[0122] In some aspects, provided is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status.

[0123] In some aspects, provided is a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: receiving, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a bloodbased sample from an individual; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more40MF-363846971Attorney Docket No.: 347372000140 processors, the end-motif sequence information into a trained machine learning model to output a dementia status.

[0124] In some aspects, provided is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: obtain methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determine a dementia methylation profile based on the methylation information. In some embodiments, the one or more programs comprising instructions, which when executed by the one or more processors of the system, further cause the system to compare the dementia methylation profile to a reference profile. In some embodiments, the one or more programs comprising instructions, which when executed by the one or more processors of the system, further cause the system to determine a dementia status based on the comparison of the dementia methylation profile to the reference profile.

[0125] In some aspects, provided is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: receive, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and input, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status.

[0126] In some aspects, provided is a non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: receive, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of41MF-363846971Attorney Docket No.: 347372000140 sequence reads form the sequence read information; and input, using the one or more processors, the end-motif sequence information into a trained machine learning model to output a dementia status.EXAMPLES

[0127] The following examples are included for illustrative purposes only and are not intended to limit the scope of the invention.Example 1 - Non-Invasive Detection of Dementia using Plasma Cell-Free DNA sequencing and Artificial Intelligence

[0128] Obtaining brain samples for diagnostics and therapeutics development has traditionally been challenging. Routine biopsies are not obtained to detect or study neurodegenerative disease pathology. Furthermore, biopsies obtained are confounded by the biopsy process itself, as their viability may be limited. As a result, brain studies are biased toward post-mortem samples, which offer insight into advanced disease but cannot effectively generalize to the biology of early-stage disease.

[0129] In the bloodstream, there is DNA (or genetic material) that circulates, which comes from each of the cells in the body. Some of this circulating DNA (or cell-free DNA) comes from the brain. To make our blood test for dementia, we have developed computational methods for analyzing DNA sequencing data to detect DNA from the brain in plasma samples.METHODSCohort outline

[0130] Plasma was collected from 145 individuals, including individuals with a historical diagnosis of dementia or cognitively healthy controls collected in the United Kingdom and Sweden. For this study, 1 ml of plasma was collected from each individual. The metadata collected included ACE score, sex, age, and educational score.Plasma extraction42MF-363846971Attorney Docket No.: 347372000140

[0131] According to the manufacturer's instructions, up to 1 ml of plasma was extracted per patient using the Qiagen Circulating Nucleic Acid kitl 9. IOOUL proteinase K was added to each sample pre-extraction. Samples were eluted into a 50uL volume and stored at - 80degC. Plasma eluates were quantified using Tapestation.Library preparation

[0132] Each eluate (generated above) was used as input into the NEBNext Enzymatic Methyl-seq library preparation per the manufacturer’s instructions20. Libraries were quantified using qPCR, pooled, and then sequenced on Illumina NovaSeq sequencers to a mean depth of 10-20x whole-genome sequencing (WGS).Sequencing pipeline

[0133] Raw sequencing data underwent quality assessment using FastQC, and then trimming was performed using TrimGalore with the following settings: 2 colour mode, paired-end mode, and a minimum mapping quality of 20. Trimmed reads were aligned to the hg38 reference genome, which was appended with the pucl9 genome (version M77789.2) and the lambda genome (version J02459.1), which represent non-human genomes spiked to assess enzymatic methylation conversion efficiency. Bismark was used to align reads using the align SE bismark function, with the following settings: minimum mapping quality of 20, up to 20 consecutive seed extension attempts allowed, seed substring length 20, up to 20 reseed attempts, 0 mismatches allowed in a seed alignment, and 5’ and 3’ clipping set to 5 bp.

[0134] Sambamba view was used to filter out poorly aligned reads. For paired-end read filtering, the following read types were discarded: secondary alignments, failed quality control, supplementary alignments, and not in proper pairs. Duplicate reads were marked and removed using Bismark; then methylation extraction was performed with the following settings: minimum depth of 10, maximum variant fraction of 0.25, merge context = false, params. methyl kit = false, bismark.bedgraph = false. CHH conversion rate was calculated using a custom shell script. Quality control metrics were generated using PicardTools and MultiQC. Spike in reads were obtained with Samtools view for quantification of conversion efficiency.Differentially methylated region identification

[0135] Publicly available tissue methylation was utilized to identify differentially methylated regions (DMRs). These files provided genome-wide methylation profiles, which43MF-363846971Attorney Docket No.: 347372000140 were processed to identify regions with significant methylation variation across the dataset. We used wgbstools (https: / / github.com / nloyfer / wgbs_tools) fmd_markers command to identify DMRs as described by Loyfer et al.

[0136] Data were filtered before DMR calling: a minimum number of 3 CpGs per read and a minimum methylation difference of 0.3 between target and background were required for a DMR. DMRs were called using a T-test with a p-value cutoff of 0.05. An alternative configuration used a 0.5 minimum methylation difference threshold with no p-value cutoff.CHH methylation analysis

[0137] CHH methylation was called using a custom version of MethylDackel software’s perRead command. The perRead command reports the methylation status of all Cs located on the reference segment to which the read aligns. BAM files aligned to the hg38 genome were used as input. Following the above command, each BAM file was annotated with an XM field containing the per-base methylation status of each read. CHH methylation was then summarized by context using a custom Python script. A dataframe of methylation per context per DMR was generated for each sample.Context summary generation

[0138] This custom script takes the following inputs and summarizes the methylation status of each sample per context. Inputs are as follows:• BAM file with per base methylation context information encoded into XM SAM tag field.• BED file containing marker region of interest on hg38 genome. o We are using DMR found from methbank data focusing on CAC context methylation.• Integer padding (in this case, Ibp), which is the length of the context, will be extracted from a known C base.

[0139] The output file is as follows:• JSON file containing per region per each context occurrence and methylation state. Note that each area was previously determined in the DMR calling step. For each44MF-363846971Attorney Docket No.: 347372000140 region, the aggregated methylation counts for each context-specific methylation signature are shown below:{“Chr7:2656380-2656519”{“CH_AACAAA”: {“total”: 0, “meth count”: 0},“CpG GGCGGT”: {“total”: 29, “meth_count”: 28}}}

[0140] The above JSON file is then converted into a format suitable as an input for a machine learning model. The summary JSON file was loaded into pandas DataFrame, and per context, per cent methylation was calculated from total and meth count fields. Here, the context window can be varied to extract different contexts such as CAC, CG, ACAC, etc. N.B. CG is standard CpG methylation, without any additional context information. CAC is an example of a CHH methylation context, though Ibp (or multiple bases) can be appended to either end.

[0141] The input for downstream machine learning approaches is the following pandasDataFrame in Table 1, which is used for model training.Table 1 - Input dataframe format for model training45MF-363846971Attorney Docket No.: 347372000140End motif analysis

[0142] We extracted end-motif sequences from cfDNA — specific nucleotide patterns at the termini of cfDNA fragments. End motifs were extracted from sequencing reads using a custom Python script to obtain 3bp or 4bp motifs. The frequency of each unique end motif was calculated across all reads in each sample.

[0143] The end-motif frequency data were then used as input into a CatBoost-based classifier. The classifier was trained on labelled data, with motif frequencies as features. Model performance was evaluated using cross-validation and metrics such as accuracy, precision, and area under the receiver operating characteristic (AUROC) curve.Classification of methylation using a Random Forest model

[0144] A random forest classification model (sklearn.ensemble.RandomForestClassifier with default parameters) was employed to analyze and classify the data based on the identified DMR regions, using CHH (or CpG) methylation levels.

[0145] To ensure the robustness and generalizability of our random forest classification model, we divided the dataset into training and testing sets. The data was split with an 80 / 20 ratio, where 80% of the data was allocated for training, and the remaining 20% was reserved for testing. This test set was set aside and not used during model training to provide an unbiased evaluation of the model’s performance after training.

[0146] We implemented a 5-fold cross-validation strategy within the training set using the Python skleam package to optimize and validate the model further. The training data was split into five equal sized subsets. In each fold, the model was trained on four subsets (80% of the training data) and validated on the remaining subset (20% of the training data). This process was repeated five times, ensuring that each subset served as the validation set precisely once. Cross-validation allowed us to assess the model’s performance across different data partitions, reducing the risk of overfitting and ensuring that the model generalized well across the entire dataset.

[0147] The final training and validation metrics were averaged across the five folds to obtain robust performance estimates. These metrics were then used to fine-tune model parameters before applying the trained model to the independent test set.46MF-363846971Attorney Docket No.: 347372000140RESULTSIdentification of brain-specific DMR regions

[0148] We studied tissue and plasma sequencing data to identify specific differentially methylated regions (DMR) in dementia. We iteratively performed DMR calling on CpG, CHH and CAC (a subtype of CHH) regions. FIG. 3 A shows a heatmap of DMRs across multiple tissues.

[0149] We then used a genomic windows-based approach to manually identify DMRs in plasma, whereby the genome was split into lOOkb windows. Here, to better characterize the signal of each methylation class, Ibp of additional context was added to each type of methylation. For each window, a T-test was used to identify differential levels of methylation of each type, and the counts of each were aggregated (See FIG. 3B). This showed varying levels of differential methylation across each type of methylation, and also that this signal varied by context.

[0150] Based on the location of CAC DMRs identified in brain tissue, we identified putative genes that are preferentially expressed in the brain sequencing (p = 0.054). Of these markers, 13 (81%) are in genic regions, 1 (6%) overlap exons, and 4 of these markers overlap genes preferentially expressed in the brain, based on public tissue expression data.Detection of Alzheimer ’s disease using CpG / CHH methylation with a context-based approach

[0151] By applying machine learning-based classifiers to filtered data, we developed a number of methods to characterize generated data. Using cfDNA methylation data for each sample, we generated a context-based methylation summary for CpG and CHH methylation. We used a context-based approach based on earlier results, which demonstrated variation in noise by context, in addition to evidence to suggest that specific contexts are abundant in the brain (See FIGS. 3A and 3B).

[0152] CHH methylation was called using a custom version of MethylDackel software perRead command, then summarized by context using a custom Python script. A dataframe of methylation per context per DMR was generated for each sample, which was used as input47MF-363846971Attorney Docket No.: 347372000140 into a Random Forest classifier. To assess the performance of this approach, data were split into train, test and validation sets.

[0153] The ROC AUC values are shown in FIGS. 4A-4C for CHH and CpG methylation, including with different methylation sequence contexts, using neuronal regionspecific DMRs. In FIG. 4B, the top performing context was found to be ACAC (AUC of 0.79). The ACAC context outperformed the CAC context (AUC of 0.50) in the validation dataset, suggesting that while signal may be present in mCAC fragments, the effect of context-specific noise may play an important role in its classification / detection performance.Detection of AD using fragmentomic approach

[0154] The study of fragmentation patterns of cell-free DNA (cfDNA), such as fragment length, fragment end-motifs, and fragment start-end positions, is known as fragmentomics and has been demonstrated in oncology previously. We hypothesized that cfDNA fragmentation profiles might be used to detect neurodegenerative diseases such as AD.

[0155] We studied end-motif sequences of 3bp or 4bp at the end termini of cfDNA fragments using a custom script. The most abundant motif was TTT, with an abundance of 16.75%. The abundance of each motif per sample was used as input into a tree-based machine-learning model. Using such an approach, the classification performance for case vs. control was 0.78 (See FIG. 4D). This generalized to an independent cohort of individuals with vs. without AD, with an AUC of 0.82.

[0156] The present invention is not intended to be limited in scope to the particular disclosed embodiments, which are provided, for example, to illustrate various aspects of the invention. Various modifications to the disclosure provided herein will become apparent from the description and teachings herein. Such variations may be practiced without departing from the true scope and spirit of the disclosure and are intended to fall within the scope of the present disclosure.48MF-363846971

Claims

Attorney Docket No.: 347372000140CLAIMSWhat is claimed is:

1. A method of determining a dementia methylation profile from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determining the dementia methylation profile based on the methylation information.

2. The method of claim 1, further comprising comparing the dementia methylation profile to a reference profile.

3. The method of claim 2, further comprising determining a dementia status based on the comparison of the dementia methylation profile to the reference profile.

4. The method of claim 3, wherein the dementia status indicates that the individual has a dementia.

5. A method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of the cfDNA; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non- CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output the dementia status.49MF-363846971Attorney Docket No.: 3473720001406. The method of any one of claims 1-5, wherein the one or more non-CpG methylation loci comprises one or more CHH methylation loci and / or one or more CHG methylation loci, where H is any of A, T, or C.

7. The method of claim 6, wherein each of the one or more CHH methylation loci is 5'-CAA-3', 5'-CAT-3', 5'-CAC-3', 5'-CTA-3', 5'-CTT-3', 5'-CTC-3', 5'-CCA-3', 5'- CCT-3', or 5'-CCC-3'.

8. The method of claim 6 or 7, wherein each of the one or more CHG methylation loci is 5 -CAG-3', 5'-CTG-3', or 5'-CCG-3'.

9. The method of claim 6, wherein the one or more methylation loci comprises 5'- CAC-3'.

10. The method of any one of claims 1-9, wherein the methylation information comprises the methylation state of each or a subset of the one or more non-CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs.

11. The method of claim 10, wherein the methylation information comprises a percent methylated for each or the subset of the one or more non-CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs.

12. The method of any one of claims 1-11, wherein the methylation information further comprises information from the methylation-processed sequence read information of one or more CpG methylation loci mapping to the one or more DMRs.

13. The method of claim 12, wherein the methylation information comprises the methylation state of each of the one or more CpG methylation loci from the methylation-processed sequence read information mapping to the one or more DMRs.

14. The method of claim 12 or 13, wherein the methylation information comprises a percent methylated for each of the one or more CpG methylation loci from the50MF-363846971Attorney Docket No.: 347372000140 methylation-processed sequence read information mapping to the one or more DMRs.

15. The method of any one of claims 1-14, wherein the methylation information is compiled in a text format.

16. The method of any one of claims 1-15, wherein the methylation information is compiled into a two-dimensional data structure.

17. The method of claim 16, wherein the two-dimensional data structure is a dataframe.

18. The method of any one of claims 1-17, wherein the methylation information further comprises sequence context.

19. The method of claim 18, wherein the sequence context is one or more bases on the 5' and / or 3' side a methylation locus.

20. The method of any one of claims 1-19, wherein the methylation information comprises information of a plurality of non-CpG locus types in the one or more DMRs.

21. The method of any one of claims 1-19, wherein the methylation information comprises information for one non-CpG locus type in the one or more DMRs.

22. The method of any one of claims 18-21, wherein the methylation information comprises information for a plurality of sequence context types in the one or more DMRs.

23. The method of any one of claims 18-21, wherein the methylation information comprises information for one sequence context type in the one or more DMRs.

24. The method of any one of claims 1-23, further comprising obtaining the one or more DMRs.

25. The method of any one of claims 1-24, further comprising determining the one or more DMRs.51MF-363846971Attorney Docket No.: 34737200014026. The method of claim 25, wherein the determining the one or more DMRs is based on one or more of CpG methylation loci, nil context, one or more non-CpG methylation loci, sequence context, one or more non-CpG methylation loci with sequence context, a tissue type, a cell type, or a disease.

27. The method of claim 26, wherein the tissue type is a brain tissue.

28. The method of claim 27, wherein the brain tissue is the frontal lobe, parietal lobe, temporal lobe, occipital lobe, insula, hippocampus, amygdala, thalamus, hypothalamus, epithalamus, subthalamus, basal ganglia, caudate nucleus, putamen, globus pallidus, nucleus accumbens, substantia nigra, red nucleus, superior colliculus, inferior colliculus, cerebellum, pons, medulla oblongata, midbrain, corpus callosum, cingulate cortex, prefrontal cortex, motor cortex, somatosensory cortex, visual cortex, auditory cortex, brainstem, pituitary gland, or pineal gland..

29. The method of any one of claims 26-28, wherein the cell type is a brain cell.

30. The method of claim 29, wherein the brain cell is a neuron, oligodendrocyte, astrocyte, pericyte, neural stem cell, or microglia.

31. The method of any one of claims 26-30, wherein the disease is Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, amyotrophic lateral sclerosis (ALS), frontotemporal dementia (FTD), Lewy body dementia, vascular dementia, corticobasal degeneration, progressive supranuclear palsy, multiple system atrophy, spinocerebellar ataxia, Friedreich’s ataxia, prion disease (including Creutzfeldt-Jakob disease), spinal muscular atrophy, Batten disease, Niemann- Pick disease, metachromatic leukodystrophy, Wilson’s disease, Alzheimer’s disease with cerebral amyloid angiopathy, or posterior cortical atrophy.

32. A method for determining a dementia status from cell-free DNA (cfDNA) from a blood-based sample obtained from an individual, the method comprising: receiving, at one or more processors, sequence read information from a sequencing analysis of the cfDNA;52MF-363846971Attorney Docket No.: 347372000140 extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status.

33. The method of claim 32, wherein the 5'- and / or 3' terminal end motifs comprise at least 2 bases in length.

34. The method of claim 32 or 33, wherein the 5'- and / or 3' terminal end motifs are 3 bases in length.

35. The method of claim 32 or 33, wherein the 5'- and / or 3' terminal end motifs are 4 bases in length.

36. The method of any one of claims 32-35, wherein the information regarding 5'- and / or 3' terminal end motifs is the frequency of each unique end motif.

37. The method of any one of claims 5-36, wherein the trained machine learning model comprises a classifier.

38. The method of any one of claims 5-37, wherein the trained machine learning model comprises a tree-based classifier.

39. The method of any one of claims 5-38, wherein the trained machine learning model is an ensemble classifier.

40. The method of any one of claims 5-39, wherein the trained machine learning model comprises a gradient boosting algorithm.

41. The method of any one of claims 5-40, wherein the trained machine learning model comprises a CatBoost algorithm.

42. The method of any one of claims 5-40, wherein the trained machine learning model comprises a Random Forest classifier.53MF-363846971Attorney Docket No.: 34737200014043. The method of any one of claims 5-42, further comprising training a machine learning model to obtain the trained machine learning model.

44. The method of any one of claims 1-43, wherein the method further comprises obtaining a blood, plasma, cerebral spinal fluid (CSF), or serum sample from the individual.

45. The method of any one of claims 1-44, further comprising performing the sequencing analysis on the cfDNA.

46. The method of claim 45, wherein the sequencing analysis is performed using a methylation-processed sequencing analysis.

47. The method of claim 45 or 46, wherein the sequencing analysis is an enzymebased methylation sequencing technique.

48. The method of claim 47, wherein the enzyme-based methylation sequencing technique is EM-SEQ™.

49. The method of any one of claims 1-48, wherein the sequencing analysis is performed to a depth of about 5x to about lOOOx.

50. The method of any one of claims 1-49, wherein the sequencing analysis is a singleend mode sequencing technique or a paired-end mode sequencing technique.

51. The method of any one of claims 1-50, wherein the sequencing analysis is a whole-genome sequencing analysis.

52. The method of any one of claims 1-50, wherein the sequencing analysis is a targeted sequencing analysis.

53. The method of any one of claims 1-52, wherein, following receiving the sequence read information, the method further comprises one or more steps of quality control or sequence trimming.

54. The method of any one of claims 1-53, wherein, following receiving the sequence read information, the method further comprises mapping to a reference genome.54MF-363846971Attorney Docket No.: 34737200014055. The method of any one of claims 5-54, wherein the dementia status indicates the individual has the dementia.

56. The method of claim 55, wherein the dementia is Alzheimer’s disease or Parkinson’s disease.

57. The method of claim 55 or 56, wherein the dementia is of a pre-symptomatic stage, early stage, moderate stage, or advanced stage.

58. A method of diagnosing a dementia in an individual, the method comprising: obtaining methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting methylation information from the sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); determining a dementia methylation profile based on the methylation information; and comparing the dementia methylation profile to a reference profile to determining a dementia status thereby diagnosing the dementia in the individual.

59. A method of diagnosing a dementia in an individual, the method comprising: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non- CpG methylation loci in one or more differentially methylated regions (DMRs); and inputting, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status thereby diagnosing the dementia in the individual.

60. A method of diagnosing a dementia in an individual, the method comprising:55MF-363846971Attorney Docket No.: 347372000140 receiving, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a blood-based sample from the individual; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output the dementia status thereby diagnosing the dementia in the individual.

61. The method of any one of claims 58-60, wherein the dementia status indicates that the individual has the dementia.

62. A method of treating an individual having a dementia, the method comprising: diagnosing the individual as having the dementia according to any one of claims 58-61; and administering to the individual a therapeutic agent for the dementia.

63. The method of claim 62, wherein the therapeutic agent is selected from the group consisting of a cholinesterase inhibitor, glutamate regulator, N-methyl-D-aspartate (NMDA) antagonist, acetylcholinesterase inhibitor, antidepressant, antipsychotic, PU.1 inhibitory modulator, an antibody -based therapy, a cell-based therapy, and anticonvulsant.

64. The method of claim 62, wherein the therapeutic agent is selected from the group consisting of donepezil, galantamine, rivastigmine, lecanemab, memantine, brexpiprazole, aducanumab, donanemab, lecanemab, posiphen, RASRxl902, RASRxl911, any of glycosylated-angiotensin 1-7, MW071, MW109, EHI-16, M3, LM11A-31, CT1812, BPN14770, allopregnanolone, MW151, and MW150.

65. The method of any one of claims 58-64, wherein the dementia is Alzheimer’s disease or Parkinson’s disease.

66. The method of any one of claims 58-65, wherein the dementia is of a pre- symptomatic stage, early stage, moderate stage, or advanced stage.56MF-363846971Attorney Docket No.: 34737200014067. A system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: obtaining methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extracting methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determining a dementia methylation profile based on the methylation information.

68. The system of claim 67, wherein one or more programs further include instructions for comparing the dementia methylation profile to a reference profile.

69. The system of claim 68, wherein the one or more programs further include instructions for determining a dementia status based on the comparison of the dementia methylation profile to the reference profile.

70. A system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: receiving, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extracting, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and57MF-363846971Attorney Docket No.: 347372000140 inputting, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status.

71. A system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and storing one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for: receiving, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extracting, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and inputting, using the one or more processors, the end-motif sequence information into a trained machine learning model to output a dementia status.

72. A non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: obtain methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract methylation information from the methylation-processed sequence read information for one or more non-CpG methylation loci in one or more differentially methylated regions (DMRs); and determine a dementia methylation profile based on the methylation information.

73. The non-transitory computer-readable storage medium of claim 72, wherein the one or more programs comprising instructions, which when executed by the one or58MF-363846971Attorney Docket No.: 347372000140 more processors of the system, further cause the system to compare the dementia methylation profile to a reference profile.

74. The system of claim 68, wherein the one or more programs comprising instructions, which when executed by the one or more processors of the system, further cause the system to determine a dementia status based on the comparison of the dementia methylation profile to the reference profile.

75. A non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: receive, at one or more processors, methylation-processed sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract, using the one or more processors, methylation information from the methylation-processed sequence read information for one or more non- CpG methylation loci in one or more differentially methylated regions (DMRs); and input, using the one or more processors, the methylation information into a trained machine learning model to output a dementia status.

76. A non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of a system, cause the system to: receive, at one or more processors, sequence read information from a sequencing analysis of cfDNA from a blood-based sample from an individual; extract, using the one or more processors, end-motif sequence information comprising information regarding 5'- and / or 3' terminal end motifs of sequence reads form the sequence read information; and input, using the one or more processors, the end-motif sequence information into a trained machine learning model to output a dementia status.59MF-363846971

Images