Epistatic interactions in alzheimer's disease

Abstract

The current invention provides a new classifying model for AD with a very high accuracy (94%) and a high AUC (98%). This model includes the age and the APOE ε4 allele, together with 3 VEGF-A related variants (rs34528081, rs2375981and rs7043199) involved in epistatic interactions that increase the risk for AD, 2 LSR polymorphisms (rs916147 and rs34259399) and one VEGF-A related (rs6993770) whose interactions are associated with decreased risk for AD (protective factors) and 2 VEGF-A related variants (rs2639990 and rs6921438) involved in epistatic interactions that both increased and decreased risk for AD. The identification of protective variants is very interesting given that until now the strongest genetic determinant of AD was the negative impact of the presence of ε4 allele of APOE, especially since the identified protective effect was higher or similar in magnitude to the negative effect of two APOE ε4 alleles.

Description

[0001] Epistatic interactions in Alzheimer's Disease

[0002] Background

[0003] Alzheimer's disease (AD) is a chronic, multi-factorial neurodegenerative disorder and the most common cause of dementia, characterised by mental and functional impairment [1], It is associated with accumulation of neuronal amyloid plaques and early lesions primarily in hippocampus. Although elderly age does not cause the disease, it is a risk factor for AD. It is estimated that AD prevalence doubles every 5 years in individuals over the age of 65. The economic implications of AD are becoming challenging for global healthcare systems. The World Health Organization has declared AD as a global health priority, stating that AD constitutes a growing universal public health issue with enormous consequences on both individuals and communities [2], It is projected to impact approximately 152 million individuals by the year 2050 [3],

[0004] AD severely affects the life of the patient, causing dependency, disability and subsequent fatality [2], There are two basic types of AD: a) familial or early onset AD which is responsible for about 5% of the disease incidence and b) sporadic or late onset AD that accounts for about 70% of the disease burden. Late onset AD is highly heritable, and it is etiologically heterogeneous originating from a mixture of multiple genetic and environmental risk factors [2],

[0005] Some of the genes that have been associated with the risk of sporadic AD involve the ABCA7, APOE, BINI, CD2AP, CD33, CLU, CR1, EPHA1, MS4A4A / MS4A4E / MS4A6E, PICALM, EPHA1 and SORL1 genes [4], To date, genome-wide association studies have identified more than 80 genes / loci associated with the risk of AD [5],

[0006] The most important genetic factor for AD is the Apolipoprotein E (APOE) gene. The APOE gene codes for a 35 kDa glycoprotein, the apolipoprotein E (ApoE), which is strongly expressed in the brain [6], There are three most common allelic variants in the APOE gene that alter the protein sequence leading to the formation of three different APOE isoforms: APOE2 (cysll2, cysl58), APOE3 (cysll2, argl58), and APOE4 (argll2, argl58) [6, 7] arising from 3 alleles, respectively, e2, e3 and e4. These alleles are associated with different ApoE roles [7], The e4 allele is the strongest risk factor for late-onset AD [4, 7, 8], due to its association with increased amyloid deposition and is a known risk factor for CVD [9], People with one e4 allele have a 2 to 3-fold elevated risk of developing AD, while those with two e4 alleles have about 12-fold increased risk compared to people with no e4 allele. On the other hand, the E2 allele of the APOE gene appears to display a protective role, as it is associated with reduced risk for AD [7], but it remains a risk factor for Type III hyperlipidemia

[0010] , The human APOE gene is situated on the long arm of chromosome 19ql3.1, an AD-associated zone as reported by GWAS [4], The lipolysis-stimulated lipoprotein receptor (LSR) gene is also located in the same region and encodes the lipolysis-stimulated lipoprotein receptor (LSR) which recognizes ApoE as ligand

[0011] , Seeing as it constitutes an ApoE receptor, LSR is implicated in the process of managing and maintaining lipid balance in the peripheral and central nervous system. Recently, our team identified significant epistatic interactions between numerous LSR gene single nucleotide polymorphisms (SNPs) and APOE in AD patients

[0011] ,

[0007] The vascular endothelial growth factor A (VEGF-A) is also considered as a risk factor for chronic diseases, including AD. The VEGF family is heavily implicated in angiogenic regulation, neurogenesis and neuronal survival

[0012] , Although an inverse relation has also been demonstrated

[0013] , decreased levels of VEGF-A in serum and cerebrospinal fluid have been linked with increased risk for AD and cognitive impairment [14, 15],

[0008] Ten genetic variants have been identified in two GWAS that explain more than 50% of the individual variability of VEGF-A levels [16, 17], These genetic determinants have been associated with intermediate phenotypes of CVD and other chronic diseases, such as autoimmune thyroid disease and depression [18-22], where VEGF-A is involved in several of their pathophysiology pathways.

[0009] Precise detection of people at high risk for AD is very critical for early diagnosis and appropriate management and can promote close monitoring, enhanced care, as well as the conduct of targeted risk factors-based interventions

[0023] , Even though late-onset AD is known to be a multifactorial disease with a strong genetic component, the use of common genetic variations identified in GWAS in disease prediction modelling has been of limited value so far, given that such polymorphisms explain a small relative risk and proportion of the underlying genetic contribution. The predictive ability of models would be improved with the combination of non-genetic information

[0024] , the inclusion of true functional variants and the incorporation of epistatic effects.

[0010] The role of VEGF-A-related genetic variants and their epistatic interactions with APOE and LSR polymorphisms in AD risk was presented in a previously filed application by the current inventors, PCT / EP2023 / 052660. As also cited in scientific literature

[0025] , a machine learning model was employed, i.e. Logistic Regression (LR) with an elastic net penalty, to study the relation of direct and epistatic interactions of known polymorphisms of APOE and LSR and VEGF-A related variants with AD along age and sex as input features. The present disclosure is based on a further study which used a much larger sample size and has revealed further interactions involved in determining the risk of determining Alzheimer's Disease.

[0011] References

[0012] 1. Soria Lopez JA, Gonzalez HM and Leger GC. (2019). Alzheimer's disease. In: Elsevier, ed. Handbook of Clinical Neurology Elsevier), pp. 231-255.

[0013] 2. Lane CA, Hardy J and Schott JM. Alzheimer's disease. EurJ Neurol. 2018; 25(l):59-70.

[0014] 3. Nichols E, Steinmetz JD, Vollset SE, Fukutaki K, Chalek J, Abd-Allah F, Abdoli A, Abualhasan A, Abu- Gharbieh E, Akram TT. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: An analysis for the Global Burden of Disease Study 2019. Lancet Public Health. 2022;7:el05-el25. doi: 10.1016 / 52468-2667(21)00249-8.

[0015] 4. Tanzi RE. The genetics of Alzheimer disease. Cold Spring Harb Perspect Med. 2012; 2(10).

[0016] 5. Andrews SJ, Renton AE, Fulton-Howard B, Podlesny-Drabiniok A, Marcora E, Goate AM. The complex genetic architecture of Alzheimer's disease: novel insights and future directions. EBioMedicine. 2023 Apr;90:104511.

[0017] 6. Munoz SS, Garner B and Ooi L. Understanding the Role of ApoE Fragments in Alzheimer's Disease. Neurochem Res. 2019; 44(6):1297-1305.

[0018] 7. Kim J, Basak JM and Holtzman DM. The role of apolipoprotein E in Alzheimer's disease. Neuron. 2009; 63(3):287-303.

[0019] 8. Siest G, Bertrand P, Qin B, Herbeth B, Serot JM, Masana L, Ribalta J, Passmore AP, Evans A, Ferrari M, Franceschi M, Shepherd J, Cuchel M, et al. Apolipoprotein E polymorphism and serum concentration in Alzheimer's disease in nine European centres: the ApoEurope study. ApoEurope group. Clin Chem Lab Med. 2000; 38(8):721-730.

[0020] 9. Liu S, Liu J, Weng R, Gu X and Zhong Z. Apolipoprotein E gene polymorphism and the risk of cardiovascular disease and type 2 diabetes. BMC Cardiovasc Disord. 2019; 19(1):213.

[0021] 10. Geisel J, Bunte T, Bodis M, Oette K and Herrmann W. Apolipoprotein E2 / E2 genotype in combination with mutations in the LDL receptor gene causes type III hyperlipoproteinemia. Clin Chem Lab Med. 2002; 40(5):475-479.

[0022] 11. Xie T, Akbar S, Stathopoulou MG, Oster T, Masson C, Yen FT and Visvikis-Siest S. Epistatic interaction of apolipoprotein E and lipolysis-stimulated lipoprotein receptor genetic variants is associated with Alzheimer's disease. Neurobiol Aging. 2018; 69:292. e291-292.e295.

[0023] 12. Beazley-Long N, Hua J, Jehle T, Hulse RP, Dersch R, Lehrling C, Bevan H, Qiu Y, Lagreze WA, Wynick D, Churchill AJ, Kehoe P, Harper SJ, etal. VEGF-A165b is an endogenous neuroprotective splice isoform of vascular endothelial growth factor A in vivo and in vitro. Am J Pathol. 2013; 183(3):918-929.

[0024] 13. Tarkowski E, Issa R, Sjogren M, Wallin A, Blennow K, Tarkowski A and Kumar P. Increased intrathecal levels of the angiogenic factors VEGF and TGF-beta in Alzheimer's disease and vascular dementia. Neurobiol Aging. 2002; 23(2):237-243.

[0025] 14. Hohman TJ, Bell SP and Jefferson AL. The role of vascular endothelial growth factor in neurodegeneration and cognitive decline: exploring interactions with biomarkers of Alzheimer disease. JAMA Neurol. 2015; 72(5):520-529. 15. Huang L, Jia J and Liu R. Decreased serum levels of the angiogenic factors VEGF and TGF-pi in Alzheimer's disease and amnestic mild cognitive impairment. Neurosci Lett. 2013; 550:60-63.

[0026] 16. Debette S, Visvikis-Siest S, Chen MH, Ndiaye NC, Song C, Destefano A, Safa R, Azimi Nezhad M, Sawyer D, Marteau JB, Xanthakis V, Siest G, Sullivan L, et al. Identification of cis- and trans-acting genetic variants explaining up to half the variation in circulating vascular endothelial growth factor levels. Circ Res. 2011; 109(5):554-563.

[0027] 17. Choi SH, Ruggiero D, Sorice R, Song C, Nutile T, Vernon Smith A, Concas MP, Traglia M, Barbieri C, Ndiaye NC, Stathopoulou MG, Lagou V, Maestrale GB, et al. Six Novel Loci Associated with Circulating VEGF Levels Identified by a Meta-analysis of Genome-Wide Association Studies. PLoS genetics. 2016; 12(2): el005874.

[0028] 18. Azimi-Nezhad M, Stathopoulou MG, Bonnefond A, Rancier M, Saleh A, Lamont J, Fitzgerald P, Ndiaye NC and Visvikis-Siest S. Associations of vascular endothelial growth factor (VEGF) with adhesion and inflammation molecules in a healthy population. Cytokine. 2013; 61(2):602-607.

[0029] 19. Stathopoulou MG, Bonnefond A, Ndiaye NC, Azimi-Nezhad M, El Shamieh S, Saleh A, Rancier M, Siest G, Lamont J, Fitzgerald P and Visvikis-Siest S. A common variant highly associated with plasma VEGFA levels also contributes to the variation of both LDL-C and HDL-C. Journal of lipid research. 2013; 54(2):535-541.

[0030] 20. Xie T, Stathopoulou MG, de Andres F, Siest G, Murray H, Martin M, Cobaleda J, Delgado A, Lamont J, Penas LE, A LL and Visvikis-Siest S. VEGF-related polymorphisms identified by GWAS and risk for major depression. Translational psychiatry. 2017; 7(3): el055.

[0031] 21. Rancier M, Zaaber I, Stathopoulou MG, Saleh A, Marmouch H, Masson C, Murray H, Kurth MJ, Lamont J, Fitzgerald P, Mahjoub S, Said K, Bel Hadj Jrad Tensaout B, et al. Pro- and anti-angiogenic VEGF mRNAs in autoimmune thyroid diseases. Autoimmunity. 2016; In press.

[0032] 22. Zaaber I, Rancier M, Stathopoulou MG, Saleh A, Marmouch H, Masson C, Murray H, Kurth MJ, Lamont J, Fitzgerald P, Mahjoub S, Said K, Bel Hadj Jrad Tensaout B, et al. Plasma VEGF-related polymorphisms are implied in autoimmune thyroid diseases. Autoimmunity. 2016; 49(4):229-235.

[0033] 23. Marinescu R, Oxtoby N, Young A, Bron E, Toga A, Weiner M, Barkhof F, Fox N, Golland P, Klein S and Alexander D. (2019). TADPOLE Challenge: Accurate Alzheimer's Disease Prediction Through Crowdsourced Forecasting of Future Data. Predictive Intelligence in Medicine: Springer International Publishing), pp. 1-10.

[0034] 24. Van Cauwenberghe C, Van Broeckhoven C and Sleegers K. The genetic landscape of Alzheimer disease: clinical implications and perspectives. Genet Med. 2016; 18(5):421-430.

[0035] 25. Petrelis AM, Stathopoulou MG, Kafyra M, et al. VEGF-A-related genetic variants protect against Alzheimer's disease. Aging (Albany NY). 2022;14(6):2524-2536.

[0036] 26. Zou H and Hastie T. Regularization and variable selection via the elastic net. J R Stat Soc Ser B (Stat Methadol). 2005; 67:301-320.

[0037] IT. Shim JW and Madsen JR. VEGF Signaling in Neurological Disorders. IntJ Mol Sci. 2018; 19(1).

[0038] 28. Garcia KO, Ornellas FL, Martin PK, Patti CL, Mello LE, Frussa-Filho R, Han SW and Longo BM. Therapeutic effects of the transplantation of VEGF overexpressing bone marrow mesenchymal stem cells in the hippocampus of murine model of Alzheimer's disease. Front Aging Neurosci. 2014; 6:30.

[0039] 29. Salomon-Zimri S, Glat MJ, Barhum Y, Luz I, Boehm-Cagan A, Liraz O, Ben-Zur T, Offen D and Michaelson DM. Reversal of ApoE4-Driven Brain Pathology by Vascular Endothelial Growth Factor Treatment. J Alzheimers Dis. 2016; 53(4):1443-1458. 30. Moore AM, Mahoney E, Dumitrescu L, De Jager PL, Koran MEI, Petyuk VA, Robinson RA, Ruderfer DM, Cox NJ, Schneider JA, Bennett DA, Jefferson AL and Hohman TJ. APOE e4-specific associations of VEGF gene family expression with cognitive aging and Alzheimer's disease. Neurobiol Aging. 2020; 87:18-25.

[0040] 31. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, Motyer A, Vukcevic D, Delaneau O, O'Connell J, Cortes A, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018 Oct;562(7726):203-9.

[0041] Summary of the invention

[0042] The current disclosure provides epistatic interactions which can be used in the development of a prediction model for risk of AD.

[0043] In a first aspect there is provided a method of determining an individual's risk of developing Alzheimer's Disease, said method comprising determining the genotypes of two or more SNPs selected from the list comprising the lipolysis-stimulated lipoprotein receptor (LSR) gene SNPs rs34259399 and rs916147, the vascular endothelial growth factor A (VEGF-A) gene SNP rs34528081, the very low density lipoprotein receptor (VLDLR) AS1 gene SNP rs7043199, the ZFPM2 (zinc finger protein, FOG family member 2) gene SNP rs6993770, the zinc binding alcohol dehydrogenase domain containing 2 (ZADH2) gene SNP rs2639990, an intergenic SNP between the Potassium Voltage-Gated Channel Modifier Subfamily V Member 2 (KCNV2) and VLDLR genes, rs2375981 and an intergenic SNP located between LOC100132354 (Inc-RNA) and the C6orf223 gene (encoding an uncharacterized protein), rs6921438.

[0044] In a preferred embodiment the method comprises genotyping single nucleotide polymorphisms in the vascular endothelial growth factor A (VEGF-A) and the very low density lipoprotein receptor (VLDLR) genes in a sample previously isolated from said individual, wherein the CT / CT genotype at the position of rs34528081 in combination with the TT or TA genotypes at the position of rs7043199 is associated with an increased risk of developing Alzheimer's Disease compared to a reference group, wherein the reference group are individuals who do not have this genotype combination.

[0045] In a further preferred embodiment the method comprises genotyping single nucleotide polymorphisms in the lipolysis-stimulated lipoprotein receptor (LSR) gene in a sample previously isolated from said individual, wherein the AA genotype at the position of rs916147 in combination with the GG genotype at the position of rs34259399 is associated with a decreased risk of developing Alzheimer's Disease compared to a reference group, wherein the reference group are individuals who do not have this genotype combination. The GA genotype at the position of rs916147 in combination with the GA or GG genotypes at the position rs34259399 is associated with a decreased risk of developing AD.

[0046] In a further aspect the genotype of one or more additional VEGF-A related SNPs selected from the group consisting of rs2639990, rs2375981, rs6921438 and rs6993770 can be determined in addition to those described above. In this aspect when the genotypes of rs2375981 and rs34528081 are determined the CG genotype at the position of rs2375981 in combination with the single insertion (C / CT) genotype at the position of rs34528081 is associated with an increased risk of developing AD. When the genotypes of rs34528081 and rs2639990 are determined the CT / CT genotype at the position of rs34528081 in combination with the TT genotype at the position of rs2639990 is associated with an increased risk of developing AD. When the genotypes of rs7043199 and rs2639990 are determined the TA or TT genotype at the position of rs7043199 in combination with the TT genotype at the position of rs2639990 is associated with an increased risk of developing AD. When the genotypes of rs34528081 and rs7043199 are determined the CT / CT genotype at the position of rs34528081 in combination with the TA genotype at the position of rs7043199 is associated with an increased risk of developing AD. When the genotypes of rs6921438 and rs7043199 are determined the GA genotype at the position of rs6921438 in combination with the TT genotype at the position of rs7043199 is associated with an increased risk of developing AD.

[0047] When the genotypes of rs34259399 and rs6993770 are determined, the genotype GG at the position of rs34259399 in combination with the genotype AA or AT at the position of rs6993770 is associated with a decreased risk of developing AD. When the genotypes rs34259399 and rs2639990 are determined, the genotype GG at the position of rs34259399 in combination with the genotype TT at the position of rs2639990 is associated with a decreased risk of developing AD. When the genotypes rs2639990 and rs6993770 are determined, the genotype TT at the position of rs2639990 in combination with the genotype AT at the position of rs6993770 is associated with a decreased risk of developing AD. When the genotypes rs916147 and rs6993770 are determined, the genotype GA or GG at the position of rs916147 in combination with the genotype AA at the position of rs6993770 is associated with a decreased risk of developing AD.

[0048] A further aspect comprises combining any of the interactions described above along with the number of APOE e4 alleles present or the genotypes of APOE SNPs rs429358 and rs7412, one or more additional SNPs selected from the list consisting of rsl0761741, rsll4694170, rsl740073 and rs4782371, and optionally the sex and age of subjects, into a model for the prediction of an individual's risk of developing Alzheimer's Disease. Brief description of the drawings

[0049] Figure 1 Machine Learning Model (Logistic Regression with Elastic Net Penalty) takes input features extracted from the UK Biobank and classifies each participant to either a case or a class. The input features include interactions of SNPs of APOE, LSR and VEGF-A related variants along Age and Sex.

[0050] Figure 2 Accuracy and Area under the Curve of the M L model when tested on the held-out test set.

[0051] Figure 3 Regression coefficient plot to describe the size and direction of the relationship between input features and outcome variable of the ML model.

[0052] Detailed description

[0053] Early detection of AD is important for the prognosis of the patients, especially in modifying treatments during its pre-symptomatic phase and before the pathological amyloid and tau protein accumulation and the extensive brain damage, which can have great impact. Multiple approaches have been used for the early identification of AD-linked parameters including medical assessment, as well as cognitive and imaging assays that exploit multimodal biological molecules features present in AD. Moreover, studies that use regression analyses have demonstrated the relation between AD and variables like clinical examination and cognitive test scores

[0023] , More recently, artificial intelligence approaches have been used, including supervised predictive analytics tools like support vector machines, random forests, and artificial neural networks to distinguish AD cases from controls, and for identifying people at higher risk of AD in a given period of time

[0023] , In the present study, we have used EN

[0026] , which is a supervised machine learning method, to assess the role of genetic variants of common AD biomarkers on the prediction of AD risk.

[0054] Single nucleotide polymorphisms are referred to herein by their reference SNP (rs) number, searchable on the National Center for Biotechnology Information (NCBI) SNP database (https: / / www.ncbi.nlm.nih.gov / snp / ).

[0055] The novel results of the present disclosure include epistatic interactions between two LSR variants (rs916147 and rs34259399) and two VEGF-A-related variants (rs6993770 and rs2639990) which were found to be associated with a decreased risk for AD and epistatic interactions between a VEGF-A variant (rs34528081) and four VEGF-A-related variants (rs7043199, rs2375981, rs2639990 and rs6921438) which were found to be associated with an increased risk for AD.

[0056] VEGF-A has been proposed as a promising novel therapeutic approach for AD

[0027] , while in AD mice model treatment with VEGF-A leads to improvement of cognitive function

[0028] , Furthermore, higher VEGF-A concentration in the cerebrospinal fluid has been associated with slower cognitive decline in patients at risk of AD

[0013] , Thus, VEGF-A could be considered as a protective factor for AD.

[0057] Concerning the interactions between APOE and VEGF-A, it has been shown that VEGF-A exerts a neuroprotective effect in humanized APOE e4 mice and treatment with VEGF-A leads to improvements of behavioural deficits

[0029] , A recent study has demonstrated that APOE e4 interacts with VEGF-A gene expression in the brain to affect cognitive performance

[0030] , Therefore, VEGF-A alone or in interaction with APOE, seems to play an important role in AD risk and the results of the current study are in line with this notion.

[0058] Table 1 - Epistatic interactions associated with decreased risk of Alzheimer's Disease

[0059] Table 2 - Epistatic interactions associated with increased risk of Alzheimer's Disease Tables 1 and 2 show the interactions which had the strongest protective effect and those which were associated with the greatest increased risk of AD.

[0060] The LSR polymorphism rs916147 was involved in interactions with one other LSR SNP (rs34259399) and one ZFPM2 SNP (rs6993770).

[0061] The LSR polymorphism rs34259399 was involved in interactions with one other LSR SNP (rs916147) and two VEGF-A-related SNPs (rs6993770 and rs2639990).

[0062] The VEGF-A polymorphism rs34528081 was involved in interactions with three VEGF-A-related SNPs (rs7043199, rs2375981 and rs2639990).

[0063] The VLDLR polymorphism rs7043199 was involved in interactions with one VEGF-A SNP (rs34528081) and two VEGF-A-related SNPs (rs2639990 and rs6921438).

[0064] The rs7043199 SNP in an intronic variant of VLDLR-AS1 gene and is located close to the VLDLR gene and its A allele has been associated with decreased VEGF-A levels [16, 17], The VLDL receptor is a member of the low-density lipoprotein receptor family and binds ApoE. It is involved in pathways essential for the development of laminated structures and for the synaptic plasticity of the brain and is, thus, considered as a receptor that could be involved in the development of AD

[0052] , The interaction between the TT genotype of rs7043199 and the CT / CT genotype of rs34528081 presented the highest coefficient, thus was the strongest risk factor for AD in the current models.

[0065] The polymorphism rs6993770 was involved in two epistatic interactions with LSR SNPs and one interaction with a VEGF-A-related SNP. rs6993770 constitutes an intronic variant of the ZFPM2 (zinc finger protein, FOG family member 2) gene. The latter codes for the FOG family member 2, which is linked with repression of GATA mediated transcriptional activation [50, 51] and thus with haematopoiesis. The T minor allele has been associated with decreased VEGF-A levels [16, 17],

[0066] Concerning rs34528081, this SNP is an insertion and genotypes are defined herein as: CT / CT (carrier of two copies of the thymine insertion), C / CT (carrier of 1 copy of the thymine insertion), C / C (carrier of no copies of the thymine insertion). These genotypes may also be referred to as T / T, - / T and - / -.

[0067] The present prediction model was developed using the EN method, which is a machine learning approach more powerful than classical statistics methodologies. Furthermore, the identified model has a sufficient accuracy of 0.94 and an AUC of 0.98.

[0068] The prediction models presented herein consist of 8 epistatic interactions that, in combination with the with APOE e4 allele, directly predict the risk for AD. These interactions involve 9 polymorphisms in 7 genes: VLDLR-AS1, KCNV2, ZADH2, C6orf223, LSR, ZFPM2, and VEGF-A. The IPA analysis highlighted relationships between the identified genes and neurological diseases, within the first top five disorders associated with said genes, including cardiovascular disease and cancer. This finding indicates that the genetic determinants of the selected biomarkers (VEGF-A, LSR and APOE) could act as common links between important chronic diseases. In fact, most of these genes are shown to be linked in the context of an enlarged common network, the functions of which correspond to cancer, dermatological diseases and conditions, organismal injury and abnormalities.

[0069] The current disclosure provides a method of determining an individual's risk of developing Alzheimer's Disease, said method comprising determining the genotypes of two or more SNPs selected from the list comprising the lipolysis-stimulated lipoprotein receptor (LSR) gene SNPs rs34259399 and rs916147, the vascular endothelial growth factor A (VEGF-A) gene SNP rs34528081, the very low density lipoprotein receptor (VLDLR) AS1 gene SNP rs7043199, the ZFPM2 (zinc finger protein, FOG family member 2) gene SNP rs6993770, the zinc binding alcohol dehydrogenase domain containing 2 (ZADH2) gene SNP rs2639990, an intergenic SNP between the KCNV2 and VLDLR genes, rs2375981 and an intergenic SNP located between LOC100132354 (Inc-RNA) and the C6orf223 gene (encoding an uncharacterized protein), rs6921438.

[0070] In a preferred embodiment the method comprises genotyping single nucleotide polymorphisms (SNPs) rs34528081 in the VEGF-A gene and rs7043199 in the VLDLR gene, in a sample previously isolated from said individual, wherein the presence of two thymine (T) nucleotides (CT / CT genotype) at the position of rs34528081 in combination with the presence of one or more thymine (T) nucleotides (TA or TT genotypes) at the position of rs7043199 is associated with an increased risk of developing AD compared to a reference group, wherein the reference group consists of individuals not having this genotype combination.

[0071] A further preferred embodiment comprises genotyping single nucleotide polymorphisms (SNPs) rs916147 and rs34259399 in the LSR gene, in a sample previously isolated from said individual, wherein the presence of one or more adenine (A) nucleotides (GA or AA genotype) at the position of rs916147 in combination with the presence of one or more guanine (G) nucleotides (GA or GG genotypes) at the position of rs34259399 is associated with a decreased risk of developing AD compared to a reference group, wherein the reference group consists of individuals not having this genotype combination.

[0072] Additionally, the method of the current invention may further comprise determining the number of APOE e4 alleles present and, optionally the genotype of the SNPs rs429358, rs7412, rsl0761741, rsll4694170, rsl740073 and rs4782371. The term APOE E4 / E4 as used herein refers to a genotype with two E4 alleles. The term APOE non-e4 as used herein refers to a genotype with no E4 allele i.e. E2 / E2, E2 / E3 or E3 / E3. The term APOE E4 / non E4 refers to a heterozygous genotype with one E4 allele i.e. E2 / E4 or E3 / E4.

[0073] In one embodiment the current invention presents a prediction model for determining risk of AD for an individual, said prediction model comprising sex and age; and interactions between two or more of the following SNPs: rs34259399, rs916147, rs6993770, rs2639990, rs34528081, rs7043199, rs2375981 and rs6921438.

[0074] The novel epistatic interactions between APOE, LSR and VEGF-A related polymorphisms presented herein allow for prediction of AD risk, constituting not only a useful prediction model, but also providing new insights about molecular mechanisms that can be implicated in AD development which could have an impact as biomarkers and / or treatment targets.

[0075] The genotype of the single nucleotide polymorphisms referred to herein may be determined at the nucleic acid or protein level using any method known to those skilled in the art. Examples of assays used to detect the SNP at the nucleic acid level include but are not limited to DNA microarrays, allele specific PCR assays and Fluorescent In-situ hybridization. Examples of assays used to detect the SNP at the protein level include but are not limited to ELISA, Western blots, Immunohistochemistry and High-performance liquid chromatography. Preferably the genotype of the SNPs is determined at the nucleic acid level.

[0076] The sample used to determine the genotype of an SNP of the current invention can be any biological sample from which the SNP can be determined as known to those skilled in the art but is preferably selected from the group consisting of whole blood (herein also referred to as peripheral blood), serum, plasma, urine, saliva, tissue sample and hair. Even more preferably the sample is a blood sample. The genotyping methods of the current invention are carried out in vitro on a sample previously taken from an individual.

[0077] Methods and Results

[0078] The UK Biobank [1] was established in 2006 in order to advance scientific knowledge regarding chronic diseases, genetic disorders, and environmental influences on human health. The UK Biobank provides access upon request to genetic, hospital, GP and lifestyle data from approximately half a million British people aged 40-69 years old. Its primary objective is to facilitate research efforts aimed at improving human health and well-being by exploring complex relationships between genotype, environment, and phenotypes. Through this initiative, the UK Biobank seeks to share a comprehensive database comprising physical measurements, medical histories, and DNA sequences, to promote collaborative research for societal benefits.

[0079] Leveraging the UK Biobank's linked electronic health records, we accessed the hospital records to define the study population. Hospital records were searched for instances of AD, using the ICD-10 code G30 corresponding to the condition being investigated. Each participant's diagnosis was confirmed through careful review of their hospital records, while potential false positives were excluded. A final set of eligible cases was established after completing this process. A set of controls without AD code was also chosen during this stage. Overall, a robust cohort of cases and controls was created via this approach, enabling further analysis towards better understanding the genetic underpinnings of AD. Using detailed hospital records helped accurately define the study population and minimise the effects of confounding factors, thus enhancing the validity and reliability of the resulting conclusions.

[0080] In this research, the dataset included a total sample size of 2719 AD patients, consisting of 1405 females and 1314 males, with a mean age of 76.71 ± 4.30 years. Within the pool of 446,892 controls, we randomly selected 2719 controls; 1463 females and 1256 males (mean age: 57.26 ± 8.09 years) to avoid the case / control imbalance. The demographic breakdown reflects the diversity of the AD patient population and the broader general population, highlighting the significance of considering gender and age when analysing cognitive decline and neurodegenerative processes. All individuals included in this study were of genetically determined European ancestry.

[0081] In the above-described population sample, we specifically examined genomic variants associated with genes APOE (SNPs: rs7412, rs429358), LSR (SNPs: rs916147, rs34259399), and 10 single nucleotide polymorphisms (SNPs: rsl0761741, rsl0738760, rs6921438, rs7043199, rs6993770, rs4416670, rsll4694170, rs34528081, rs4782371 and rs2639990) related to vascular endothelial growth factor-A (VEGF-A).

[0082] Each SNP genotype was assigned a categorical value either 0 (two alternate alleles), 1 (an alternate and a reference alleles) or 2 (two reference alleles). To consider both independent and interaction effects among these SNPs, all possible pair of combinations of the SNPs were considered as input features. Specifically, the presence of certain genetic variations related to the APOE gene (E2, E3, and E4 alleles) was assessed in relation to two specific SNPs (rs429358 and rs7412). Based on the combinations of these genetic variations, new binary variables were created in the dataset. For instance, if an individual exhibited a particular combination of alleles (e.g., rs429358_C = 2 and rs7412_T = 0), a corresponding binary variable (APOE_E4_E4) was assigned a value of 1; otherwise, it was assigned a value of 0. Similar conditions were applied to define other allele combinations, resulting in the creation of binary variables representing different genetic profiles. Finally, composite variables were generated to represent broader genetic categories, such as APOE non-e4 carriers and APOE E4 carriers, by combining the previously defined binary variables.

[0083] Subsequently, the input features of SNP interactions alongside sex and age served as input features for LR EN machine learning classification algorithm aimed at predicting AD risk, as illustrated in Figure 1.

[0084] We allocated 75% of the data for training the ML model and the remaining 25% served as the test dataset (held-out test set), intended for subsequently estimating the models' generalization abilities. An equal number of non-affected samples were randomly selected to maintain balance across classes, followed by random partitioning into training and testing cohorts (allocated in a ratio of 75% to 25%, respectively). The final model demonstrated high performance metrics upon evaluation against the held-out test dataset containing confirmed instances of AD drawn from the UK Biobank.

[0085] Utilizing a LR classifier equipped with an EN penalty— a method combining LI and L2 regularization schemes derived from lasso and ridge regression methods, respectively— these input attributes were processed. Through employing k-fold cross-validation and grid search methodologies to optimize the hyper-parameters within this algorithm, a robust predictive tool emerged. During our training, we utilized LR classifiers implemented via Scikit-learn's GridSearchCV module for hyper-parameter tuning. Specifically, we explored four different settings for maximumjteration, ranging from 300 to 10000 iterations, ensuring convergence of algorithm. Furthermore, we examined elastic net regularization by adjusting the LI ratio parameter across four discrete value intervals (0.4, 0.5, 0.6, and 0.7) in order to balance feature selection sensitivity and stability. For the penalty, we applied EN exclusively, employing SAGA solvers known for their efficiency and accuracy in LR tasks. Lastly, we executed cross-validation with five splits to ensure the generalizability of the models' performances while scoring them solely based on accuracy metrics. The optimal combination of these hyperparameters was determined automatically by the GridSearchCV procedure, fitting the LR classifier to the training dataset comprising features based on sex, age and SNP interactions. This approach allowed us to discover an effective configuration that maximizes classification performance while preventing over-fitting concerns. Results

[0086] Table 3 presents the minor allele frequency (MAF) for each selected SNP in the whole UK Biobank. Each row corresponds to a different SNP, with the first column indicating the SNP identifier (SNP) and the last column indicating its corresponding MAF value.

[0087] Table 3 - Minor Allele Frequency (MAF) of Selected Single Nucleotide Polymorphisms (SNPs)

[0088] *reference allele may also be referred to as C to denote no insertion and the alternate allele as CT to denote the presence of a T insertion

[0089] After evaluation of the LR classifier on the test dataset, we achieved a remarkable prediction accuracy score of 0.94, reflecting the proficiency of the trained model in discerning positive cases from negatives ones. Utilizing the Area Under the Curve (AUC)-Receiver Operating Characteristics (ROC) metric, which captures the trade-off between sensitivity and specificity, yielded outcome of 0.98. As shown in Figure 2, the accompanying 95% confidence interval ranged from 0.93 to 0.97, indicating exceptional precision and narrow uncertainty bounds around the estimated ROC-AUC value with a significant p- value of 2e-16. Collectively, these compelling results demonstrate the successful application of the chosen model and highlight the significant contributions of the selected variables toward enhancing predictive performance.

[0090] We also present a regression coefficient plot elucidating the strength and orientation of relationships among input features and the resulting output variable within our chosen LR EN model. Specifically, Figure 3 depicts the impact of individual predictors upon the dependent variable, Alzheimer's disease, in this case, allowing for the interpretation of their relative significance. Our analysis included factors such as age, sex, and genetic variations represented by SNPs (Single Nucleotide Polymorphisms), taking into account all possible combinations of three genotype states: heterozygous, homozygous reference (Ref / Ref), and homozygous alternative alleles (Alt / Alt).

[0091] Consequently, this generated a total of 281 distinct input features. Upon evaluating the outcomes, we highlight the top 20 significant features, consisting of 10 with the highest positive coefficients and another 10 exhibiting the least negative coefficients.

[0092] In our analysis, we discovered that specific SNPs connected to APOE, LSR, and VEGF-A related variants are consistently associated with the risk of AD. Carriers of two copies of the E4 variant of the APOE gene face a substantially higher risk of developing AD compared to non-carriers. As indicated by its relatively large positive coefficient, this genotype strongly influences the probability of Alzheimer's diagnosis. Our findings indicated that individuals with the specific combination of alleles at these two SNPs, rs34528081_CT / CT*rs7043199_TT also have an increased likelihood of AD.

[0093] APOE_e4_non_e4, coefficient suggests that individuals carrying at least one APOE_EA alongside a non- E4 variant allele have a slightly elevated risk of AD compared to those without any APOE_EA alleles. Even though the latter has a smaller magnitude than APOP_E4_E4, it indicates that possessing just one E4 allele already raises the likelihood of AD, emphasizing the importance of considering even subtle genetic differences in assessing risks associated with this disorder.

[0094] Individuals with this specific combination of alleles at these SNPs, rs34528081_C / TC*rs2375981_CG, have a moderate increase in AD risk, although lower than some other genetic factors. Rs34528081_CT / CT*rs2639990_TT, similar to the previous SNPs interaction, was also associated with a moderate increase in AD risk. Individuals with the specific allele combination of the SNPs, rs2639990_TT*rs7043199_TT had a slightly elevated risk of AD.

[0095] Age has a positive coefficient, indicating that as age increases, the likelihood of AD diagnosis also increases. However, its effect is relatively modest compared to the above-described genetic factors. Rs34528081_CT / CT*rs7043199_TA interaction and the interaction rs6921438_GA*rs7043199_TT also had a modest increase in AD risk.

[0096] The negative coefficients observed for the interactions rs6921438_AA*rs34259399_GA, rs916147_GG*rs6993770_AA, rs916147_GA*rs6993770_AA, rs2639990_TT*rs6993770_AT, rs34259399_GG*rs2639990_TT, rs34259399_GG*rs6993770_AT, rs34259399_GG*rs6993770_AA, rs916147_GA*rs34259399_GG, rs916147_GA*rs34259399_GA and rs916147_AA*rs34259399_GG suggest their association with a decreased likelihood of AD.

Claims

Claims1. A method of determining an individual's risk of developing Alzheimer's Disease, said method comprising determining the genotypes of two or more single nucleotide polymorphisms (SNPs) selected from the list comprising the lipolysis-stimulated lipoprotein receptor (LSR) gene SNPs rs34259399 and rs916147, the vascular endothelial growth factor A ( VEGF-A) gene SNP rs34528081, the very low density lipoprotein receptor (VLDLR) AS1 gene SNP rs7043199, the zinc finger protein, FOG family member 2 (ZFPM2) gene SNP rs6993770, the zinc binding alcohol dehydrogenase domain containing 2 (ZADH2) gene SNP rs2639990, an intergenic SNP between the KCNV2 and VLDLR genes, rs2375981 and an intergenic SNP located between LOC100132354 (Inc-RNA) and the C6orf223 gene (encoding an uncharacterized protein), rs6921438.

2. The method of claim 1 comprising genotyping SNPs rs34528081 in the VEGF-A gene and rs7043199 in the VLDLR gene, in a sample previously isolated from said individual, wherein the presence of two thymine (T) nucleotides (CT / CT genotype) at the position of rs34528081 in combination with the presence of one or more thymine (T) nucleotides (TA or TT genotypes) at the position of rs7043199 is associated with an increased risk of developing AD compared to a reference group, wherein the reference group consists of individuals not having this genotype combination.

3. The method of claim 2 further comprising determining the genotype of one or more additional SNPs selected from the group consisting of rs2375981, rs2639990 and rs6921438.

4. The method of claim 3 wherein when the genotypes of rs34528081 and rs2375981 are determined, the genotype C / CT at the position of rs34528081 in combination with the genotype CG at the position of rs2375981 is associated with a decreased risk of developing AD.

5. The method of claim 3 wherein when the genotypes of rs34528081 and rs2639990 are determined, the genotype CT / CT at the position of rs34528081 in combination with the genotype TT at the position of rs2639990 is associated with a decreased risk of developing AD.

6. The method of claim 3 wherein when the genotypes of rs2639990 and rs7043199 are determined, the genotype TT at the position of rs2639990 in combination with the genotype TT or TA at the position of rs7043199 is associated with a decreased risk of developing AD.

7. The method of claim 3 wherein when the genotypes of rs6921438 and rs7043199 are determined, the genotype GA at the position of rs6921438 in combination with the genotype TT at the position of rs7043199 is associated with a decreased risk of developing AD.

8. The method of claim 1 comprising genotyping single nucleotide polymorphisms (SNPs) rs916147 and rs34259399 in the LSR gene, in a sample previously isolated from said individual, wherein the presence of one or more adenine (A) nucleotides (GA or AA genotype) at the position of rs916147 in combination with the presence of one or more guanine (G) nucleotides (GA or GG genotypes) at the position of rs34259399 is associated with a decreased risk of developing AD compared to a reference group, wherein the reference group consists of individuals not having this genotype combination.

9. The method of claim 8 further comprising determining the genotype of one or more additional SNPs selected from the group consisting of rs6993770 and rs2639990.

10. The method of claim 9 wherein when the genotypes of rs34259399 and rs6993770 are determined, the genotype GG at the position of rs34259399 in combination with the genotype AA or AT at the position of rs6993770 is associated with a decreased risk of developing AD.

11. The method of claim 9 wherein the genotypes rs34259399 and rs2639990 are determined, the genotype GG at the position of rs34259399 in combination with the genotype TT at the position of rs2639990 is associated with a decreased risk of developing AD.

12. The method of claim 9 wherein the genotypes rs2639990 and rs6993770 are determined, the genotype TT at the position of rs2639990 in combination with the genotype AT at the position of rs6993770 is associated with a decreased risk of developing AD.

13. The method of claim 9 wherein the genotypes rs916147 and rs6993770 are determined, the genotype GA or GG at the position of rs916147 in combination with the genotype AA at the position of rs6993770 is associated with a decreased risk of developing AD14. The method of any previous claim further comprising determining the genotype of APOE e4 SNPs rs429358 and rs7412.

15. The method of any previous claim wherein the in vitro sample is a blood sample.

16. A prediction model for determining risk of AD for an individual, said prediction model comprising sex and age; and interactions between two or more of the following SNPs: rs34259399, rs916147, rs6993770, rs2639990, rs34528081, rs7043199, rs2375981 and rs6921438.

17. The prediction model of claim 16 which also includes one or more SNPs selected from the list consisting of rs429358, rs7412, rsl0761741, rsll4694170, rsl740073 and rs4782371.

18. A method for determining if an individual is suitable for treatment with VEGF-A, said method comprising genotypes of two or more SNPs selected from the list comprising the LSR gene SNPs rs34259399 and rs916147, the VEGF-A gene SNP rs34528081, the VLDLR AS1 gene SNP rs7043199, the ZFPM2 gene SNP rs6993770, the ZADH2 gene SNP rs2639990, an intergenic SNP between the KCNV2 and VLDLR genes, rs2375981 and an intergenic SNP located between LOC100132354 (Inc-RNA) and the C6orf223 gene (encoding an uncharacterized protein), rs6921438.

Images