Delivery probe

Abstract

The present invention provides RNA depletion probes, i.e., a set of short DNA oligonucleotides that hybridize along the length of a target RNA and mediate digestion of the target RNA by RNase H, to remove excess RNA molecules from a sample. The depletion probe is designed mainly based on the biochemical and biophysical characteristics of the probe, so that all depletion probes in a group of depletion probes show a substantially uniform and consistent behavior in the aspect of binding with target RNA in a sample. The probes are designed primarily for specific performance targets and biophysical properties, resulting in probe sets having irregular, even significantly random pitches along target RNA molecules.

Description

Technical Field

[0001] The present disclosure relates to depletion probes. Background Art

[0002] The transcriptome refers to the RNA transcripts of a cell or organism at a given time. Knowledge of the transcriptome reveals active cellular processes and can provide information about cell regulation, growth, and dysfunction.

[0003] Transcriptome analysis typically involves microarray technology, and more commonly involves next generation sequencing technology (e.g., RNA-Seq). A common goal of RNA-Seq is to detect all transcripts (including mRNA and non-coding RNA) among the different transcripts present, as well as to detect splice variants, mutations, mobile genetic elements, and expression levels at various developmental stages or under various conditions. Transcriptomics and RNA-Seq are used in diagnostics, disease analysis, pathogen detection, evolutionary biology, and other research areas. For example, RNA-Seq can potentially identify genes associated with resistance to environmental stresses (such as drought resistance in crops). In another example, transcriptome analysis can provide information about the mechanisms of drug resistance, thereby potentially revealing strategies for combating hospital-acquired antibiotic-resistant infections. Summary of the invention

[0004] The present invention provides RNA depletion probes, i.e. groups of short DNA oligonucleotides, which hybridize along the length of the RNA to be depleted from a sample. The DNA oligonucleotide mediates the digestion of the target RNA by RNase H. The present invention can be used to remove non-target RNA from a sample to promote the capture of low-abundance RNA or simply achieve a higher yield of target RNA. The depletion probes according to the present invention are designed based on the biochemical and biophysical properties of the probe / target interaction so that all depletion probes in a group of depletion probes show uniform, consistent behavior. For example, the depletion probes can be designed to have a melting temperature (Tm) within a narrow, predefined range. The probes of the present invention are designed to achieve improved performance, thereby generally producing probe groups with irregular, even apparently random spacing along the RNA to be removed from a sample.

[0005] Not affected by irregular spacing, due to reasonable biologically based design (described design models Tm and screens reference RNA sequence data to omit candidate probe sequences with off-target matches), the probe groups of the present invention are superior to the probe groups of the prior art. The probe groups of the prior art sacrifice complete deletion to achieve uniform probe binding along the RNA to be removed. Due to the restrictive assumption that the probe group must be tiled along the length of RNA with uniform spacing and melting temperature, many probes are poorly combined due to factors such as GC content, secondary structure or length. The present invention solves these problems by using a design method that works in a selected melting temperature window and reduces off-target binding.

[0006] The probes of the present invention form heteroduplexes at unpredictable, apparently random spacings along one or more RNA molecules. In fact, the probe sets of the present invention may leave long RNA gaps - in some cases, gaps up to more than about 100 bases - but are still superior to the probe sets of the prior art, which are designed to follow some simple short, regular gap patterns along the target RNA. In addition, even in highly decomposed samples, the probes of the present invention show improved depletion. Another advantage of the probe design according to the present invention is that it allows probe hybridization and RNA digestion to occur in a single, rapid step, rather than using a heating / cooling cycle, followed by a long and suboptimal digestion temperature. Because the probe sets of the present invention use probes designed to exploit the actual sequence content and biophysical properties of heteroduplex formation, RNA depletion using the probe sets of the present invention reliably removes excess RNA from the sample. Because such tools can be used to reliably remove RNA, such as ribosomal RNA and / or globin transcripts, RNA-Seq assays using the probe sets of the present invention better detect remaining and sometimes rare mRNA transcripts. Because rare genetic events are markers for critical conditions such as cancer or pathogenic infections, the probe sets of the present invention can be used to detect critical biological information early and accurately.

[0007] In some aspects, the invention provides a method for depleting RNA. A preferred method includes hybridizing a plurality of DNA oligonucleotides with a target RNA molecule in a sample to form a heteroduplex, wherein the DNA oligonucleotide has a sequence selected to achieve the following: (i) the melting temperature of the heteroduplex is within a predetermined range, and (ii) the matching with reference to off-target RNA is minimized. The method of the present invention further includes digesting the RNA in the heteroduplex. The heteroduplex can be formed along the RNA molecule with a non-uniform distribution. The DNA oligonucleotide can include a low GC oligonucleotide and a high GC oligonucleotide having a melting temperature higher than the low GC oligonucleotide. The DNA oligonucleotide can include a low GC oligonucleotide and a high GC oligonucleotide shorter than the low GC oligonucleotide. The DNA oligonucleotide can be selected to minimize the spacing between adjacent heteroduplexes, and can exist with a non-uniform concentration.

[0008] In some embodiments, DNA oligonucleotides are selected by generating a set of candidate sequences complementary to the target RNA (in this case, the RNA to be removed), wherein the predicted melting temperature is within a predetermined range; assigning a cost function to each of the candidate sequences based at least in part on the match score with the reference off-target RNA; and selecting a set of candidate sequences mapped to positions along the RNA molecule, wherein the group is selected by an algorithm that minimizes the cumulative cost function, gaps between positions, and / or overlaps. The DNA sequences can be selected by a computer system to provide DNA oligonucleotides that uniquely anneal RNA molecules with similar annealing temperatures, with minimal or no overlap, and with minimal or no annealing off-target RNAs. Optionally, the computer system further selects sequences to provide DNA oligonucleotides with minimal length and / or minimize gaps between heteroduplexes.

[0009] In certain embodiments, the DNA oligonucleotide is tested and shown not to stably hybridize with the protein coding transcript in the standardized human reference RNA extract. The target RNA molecule may include any one of ribosomal RNA, globin transcript or mitochondrial RNA. The preferred method includes performing the steps to digest the ribosomal RNA and globin transcript in the sample. The digestion step may include treating the sample with RNAseH. In addition, the preferred method may include reverse transcription of non-target RNA molecules in the sample after the digestion step. In a preferred embodiment, heteroduplexes are formed at positions along the RNA molecules, and the positions do not form any regular, repeated or uniform patterns or spacings. For example, the spacing between heteroduplexes may be highly variable and / mathematically random. The positioning of the heteroduplexes may leave one or more uncovered regions of the RNA molecule having at least about 10 to about 100 bases. In addition, coverage does not need to be end-to-end, which means that 5' and 3' end sequences may be uncovered.

[0010] The method of the present invention can increase the ratio of poly-A tail RNA (poly-A tailed RNA) and non-coding RNA in the sample. The sequence of at least two DNA oligonucleotides in the DNA oligonucleotide is selected to hybridize with the RNA to be exhausted at the position overlapping or adjacent to each other. The length of the DNA oligonucleotide used in the present invention can be about 20 bases to about 40 bases (for example, changing between 18 and 44, including end value). The DNA oligonucleotide can be present in the mixture with non-equimolar concentration. For example, based on empirical observations, the concentration of some probes in the probe may be enhanced, for example, doubled.

[0011] Aspects of the present invention provide a group of RNA depletion probes. An exemplary group includes multiple DNA oligonucleotides that hybridize with the RNA to be depleted to form heteroduplexes, wherein the DNA oligonucleotides have sequences selected to achieve the following: (i) the melting temperature of the heteroduplex is within a predetermined range, and (ii) the matching with the off-target RNA is minimized. Preferably, the heteroduplex is formed along the RNA molecule with a non-uniform distribution. The spacing between the heteroduplexes in the group can be highly variable and / or mathematically random. The DNA oligonucleotides can be selected to minimize the spacing between adjacent heteroduplexes. The DNA oligonucleotides can include both low GC oligonucleotides and high GC oligonucleotides (the melting temperature of the high GC oligonucleotides is higher than the low GC oligonucleotides and / or shorter than the low GC oligonucleotides). The probe set can be designed by generating a set of candidate sequences complementary to an RNA molecule, wherein the predicted melting temperature is within a predetermined range; assigning a cost function to each of the candidate sequences based at least in part on a match score with a reference off-target RNA; and selecting a set of candidate sequences mapped to positions along the RNA molecule, wherein the set is selected by an algorithm that minimizes the cumulative cost function, gaps between positions, and / or overlaps. The probe set can be designed to deplete any of ribosomal RNA, mitochondrial RNA, globin transcripts, or others. It is noteworthy that heteroduplexes can form at positions along the RNA molecule that do not form any regular, repetitive, or uniform pattern or spacing. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A probe design method according to certain embodiments is illustrated.

[0013] Figure 2 The steps of the probe selection method are shown.

[0014] Figure 3 The distribution of inter-probe gap sizes is shown.

[0015] Figure 4 is a histogram of gap sizes.

[0016] Figure 5 is a histogram of probe lengths.

[0017] Figure 6 The diagram shows the design method of Bloom filter.

[0018] Figure 7 An example of a probe design method is described.

[0019] Figure 8 Depletion of ribosomal RNA from whole blood is shown.

[0020] Fig. 9 Results for hemoglobin RNA depletion from whole blood are shown.

[0021] Fig.10 Depletion of rRNA from formalin-fixed paraffin-embedded (FFPE) tissue with 10 ng RNA is shown.

[0022] Fig.11 Depletion of rRNA from FFPE tissue with 400 ng RNA is shown.

[0023] Fig.12 Depletion of rRNA from universal mouse and rat references is shown. DETAILED DESCRIPTION

[0024] The present disclosure relates to methods for designing and using RNA depletion probes and probe sets designed according to such methods. The methods and materials disclosed herein can be used to remove different types of RNA molecules from a sample. The methods of the present invention have specific applications in transcriptomics and certain RNA-based workflows intended to identify or quantify certain RNA molecule populations. For example, it is usually the case that specific RNA molecules, such as mRNA, in a sample need to be separated. The depletion probes of the present invention can be used to deplete ribosomal catalytic RNAs (which do not encode proteins themselves) that may interfere with the detection of target mRNA in a sample, for example. As an example, one method for separating and evaluating mRNA is to use poly-A specific initiation of oligonucleotide-dT as a method to avoid rRNA molecules. However, rRNA still exists and interferes, and there are a variety of workflows that benefit from using target capture rather than hybridizing poly-T primers to mRNA poly-A tails. In addition, partially decomposed samples (such as formalin-fixed paraffin-embedded (FFPE) samples) often separate the 3'poly-A tail of the transcript from the 5' end, which makes it impossible for technologies that rely on poly-A enrichment to obtain information on the 5' end. In addition, some RNA species, such as long non-coding RNA (lncRNA), do not contain a poly-A tail. In such instances, the ability to deplete non-target RNA makes detection of the target more difficult.

[0025] The methods and compositions of the present invention provide tools for depleting any RNA species from a sample. RNA depletion according to an embodiment of the present invention is operated by using a "depletion probe" or DNA oligonucleotide designed to hybridize with the RNA molecule to be depleted. The DNA oligonucleotide and RNA anneal to form a heteroduplex. RNase H recognizes the DNA-RNA heteroduplex and cuts the RNA in the heteroduplex. Using RNase H for RNA depletion is attractive because it does not require the RNA to be retained to contain a poly-A tail, and does not focus on whether a poly-A tail exists on the RNA molecule targeted for depletion. Therefore, even arbitrary mRNA transcripts (such as globin) can be depleted by RNase H depletion. In addition, the method of the present invention allows lncRNA and mRNA enrichment even in highly decomposed samples.

[0026] According to the present invention, RNase H depletion utilizes depletion probe, i.e. short DNA oligonucleotides hybridized with the RNA to be depleted, thereby forming heteroduplex.Some conventional methods use a group of depletion probes tiled along RNA molecules.Using tiled probes is attractive, because there is the possibility that the whole target molecule is covered.Other methods have been proposed, which allow a certain spacing between probes, such as leaving the interprobe gap of some specific number of bases between each adjacent probe target pair.In addition, some depletion probe groups can include probes described as overlapping each other, which obviously provide super coverage (for example, coverage>1).In such descriptions, "overlapping" is a recognized term, but the selection of this word is misleading in the degree of its description of two probes hybridizing with a single target molecule (one of the two probes leaves an uncombined single-stranded end, and the other probe is fully combined).In actual use, each probe in the so-called overlapping probes is likely to find an independent copy of the target RNA molecule of the same type, and each copy is fully combined to a molecule. Here, the word overlapping (and as used herein) simply recognizes that the DNA oligonucleotides are designed to complementary regions, where for both oligonucleotides it is their cognate targets that overlap.

[0027] Regardless of the probe tiling strategy, i.e., whether contiguous, evenly spaced, or overlapping, existing probe design reveals a design paradigm in which a decision is made as to which coverage pattern is most appropriate and the coverage pattern is applied to cover the molecule to be depleted. The coverage pattern may indicate that the probe targets overlap, are contiguous, or have regular spacing. Once a pattern is chosen, it is applied to the molecule to design probes, which are synthesized and used for RNA depletion.

[0028] The present invention recognizes that the laboratory reagents produced by conventional probe coverage strategies impose restrictions on their use in analytical determinations. For example, due to the diversity of GC content or due to RNA secondary structure, existing depletion probe sets can only be used under very limited conditions. Moreover, the reaction temperature of optimizing most probes means that the reaction is carried out at a temperature that is too high or too low for some of the probes in the probe. Conventional probe coverage strategies are not well suited for potential sequence similarity regions between the RNA to be depleted and the RNA of interest.

[0029] The present invention addresses such limitations. The present invention provides a method of probe design that first relies on the biophysical properties of the molecules to be deleted and the molecules to be retained. The method of the present invention uses sequence data of RNA molecules and also uses reference RNA data and applies modeling to predict probes that hybridize to the target RNA without marking the destruction of the mRNA to be retained.

[0030] Probes are designed based on biophysical properties. The methods and compositions of the present invention use sequence information, GC abundance, reference sequences and modeled / predicted biochemical behavior to generate and select probe groups that perform well, have uniform, consistent hybridization and do not have off-target hybridization in the entire probe group. The important insight of the present invention is that probe design does not need to be projected indiscriminately onto the spatial pattern on the RNA molecule length. On the contrary, establish desired properties such as the range of melting temperatures, and then design probe groups with those desired properties. For example, a tuple including all possible probes with desired properties can be generated algorithmically (for example, from sequence data), and then the tuple can be reduced based on mapping, filtering and / or scoring to select smaller but useful probe groups. The group of selected useful probes is synthesized as DNA oligonucleotides, and is used to deplete the selected RNA from the sample, so the RNA of interest is enriched. It is noteworthy that the reference data set of non-target RNA sequences can be used for, for example, filtering or scoring steps to remove probes from the tuple that will deplete the mRNA of interest that may be scarce from the sample.

[0031] In an initial step of generating all or a large number of possible probes, some embodiments start with establishing a nearly uniform Tm (or range of Tms) between any probe and its target, and also aim to minimize the Tm between each probe and any potential off-target.

[0032] A variety of methods can be used to obtain a file set. The selected probe is preferably combined across target RNA. The algorithm can be used to seek probe groups that combine across all regions of a target (e.g., 45S, mitochondrial rRNA, or globin transcript). Some embodiments use an algorithm based on a graph, and the algorithm seeks to minimize (without specific constraints) the spacing between probes and the potential Tm with off-target (e.g., possibly assessed by the homology percentage with the strongest non-target BLAST hit). Other target spanning algorithms can be used to obtain a group of probes. Probe groups can partially supplement other one or more rounds of similar designs, and / or supplement manually based on empirical data. For example, the concentration of some probes can be "enhanced" to improve performance, such as the result of an initial test assay.

[0033] Using such methods, the present invention provides methods for preparing depletion probes, depletion probe groups, and methods for using depletion probes to deplete rRNA and other highly expressed RNAs from, for example, RNA-Seq libraries based on RNaseH. Any suitable analysis package or algorithm can be used for probe design. Preferred embodiments generate candidate probes in a manner informed by the sequence information of the target and a group of reference RNA sequence information of the sample or organism (interestingly, whether the algorithm uses a reference RNA sequence data set that also includes the target is not important). The present disclosure provides probe design methods and probe groups, which are designed by a method that is systematic and reasonable and allows rapid design iterations to optimize the performance of the probe group.

[0034] At least three basic approaches to designing probes are considered. The first approach uses a Bloom filter to identify unique k-mers (Bloom filtered). The second approach generates a set of probes with equilibrium melting temperatures (Tm), which are filtered by a basic local alignment search tool (e.g., BLAST or related algorithms) to avoid generating probes that deplete non-target RNAs (BLAST filtered). The third approach generates a set of probes with equilibrium melting temperatures (Tm), then scores the probes using BLAST, and then selects the probe set that traverses the target RNA while minimizing the score (BLAST scored).

[0035] In these methods, the uniqueness of potential probes can be screened for RefSeq human RNA (manually masked for target sequences). In certain exemplary embodiments, RNA28SNx and RNA18SNx are used as developmental models. Probes for other RNAs (such as 5S and mitochondrial rRNA probes) can be designed similarly. After generating the initial probe set, the final candidate group covering the target can be selected, for example, using bioinformatics programming techniques known to those skilled in the art. The preferred embodiment adopts the BLAST scoring method.

[0036] Figure 1The BLAST scoring method 101 for probe design according to certain embodiments is illustrated. The method starts with selecting a target and obtaining sequence data. This can be performed by accessing a publicly available genetic information database (such as GenBank or Ensemble). For example, method 101 can be used to deplete Homo sapiens 18S ribosomal N3 RNA from a biological sample. The computer system of the present invention can be used to retrieve the NCBI reference sequence NR_146152.1. Software packages such as BioPerl or BioRuby can be used to import sequence data, such as the RNA18SN3 FASTA file of the 18S ribosomal RNA N2 stored in GenBank under the NCBI reference sequence NR_146152.1. This file can be brought into the system and set as a target. The system can be operated to bring in multiple targets (e.g., substantially simultaneously, such as in a "run"). A preferred embodiment sets at least two or three RNA molecules as targets, such as the longest ribosomal RNA plus globin transcripts.

[0037] After setting the target, method 101 involves setting the boundaries of the melting temperature. This can be achieved by setting the target temperature, and then allowing a range such as 50 degrees plus or minus 3. Then, the system checks the target file to generate all possible strings, and if it is synthesized as a DNA oligonucleotide, the Tm of the string will be within the boundaries. In a simple and direct embodiment, the software package on the computer system starts with the 5' end of the target file, and reads each possible k-aggregate, thereby calculating the tm of each k-aggregate. Any suitable calculation can be performed. In certain embodiments, the system uses [4(G+C)+2(A+T)]. In order to illustrate with a simple example, NR_146152.1 starts with T. Therefore, the first possible k-aggregate is a 1-aggregate composed of T. Calculation returns 2, which exceeds the boundary (not within 3 degrees of 50), so the system will not write k-aggregates to the file. Then, the system reads 5'2-aggregates (which are TA), etc. Each string generated at the Tm within the boundary is written into the file, and then the system moves to position 2 in the target file.

[0038] For this stage, it may be preferable to establish a minimum (and / or maximum) length. The system iterates over the target file and writes all possible k-mers with Tms within the bounds into the target file, for example, all probes within the bounds can be written into a FASTA file.

[0039] The method can proceed to BLAST each probe for reference RNA sequence data. In a simple and direct example, BLAST scoring is used to simply omit probes that match non-target RNAs above a certain threshold. In certain embodiments, BLAST results are used to score probes (e.g., [0..1]), wherein the scoring will be used as the cost in the cost function. The computer system operating can preferably use blastn-short.

[0040] BLAST scores can be used to filter out probes or to assign a score to a probe that is used by an algorithm that selects a set of probes to cover a target RNA.

[0041] The algorithmic task of the computer system is to cover the target RNA (for convenience, from the 5' end to the 3' end), i.e., a task that can be presented to the computer when traversing a graph in which the probes are nodes and any inter-probe gaps are edges (or simply, vice versa). The system attempts to traverse the graph using any score as a cost function for using the probes. This can be done using a version of the Dijsktra algorithm in which the edge weight becomes the "distance" between the two nodes + the "cost" for accessing the target node. In some embodiments, the edge weight includes a gap opening value and a separate gap extension value. In some other embodiments, the cost of accessing a node can be based on its similarity to a potential non-target RNA.

[0042] Figure 2 The steps of the probe selection method are illustrated. In the top inset, all probes have been generated, for example, the GenBank file is read to obtain all strings with a specific length that are calculated to have a Tm within the boundaries, and the strings are written into the FASTA file. In the figure, each rectangle is a candidate probe, i.e., a sequence from the GenBank file. The shaded blocks are probes that will be excluded by comparing BLAST with the reference RNA sequence data. The reference matching probes are excluded from consideration (directly removed in the BLAST filter, effectively avoided in the BLAST score), and the algorithm seeks to construct a directed graph through the probes. Any suitable algorithm can be used. Certain embodiments use a graph traversal algorithm, which is essentially a directed acyclic graph that seeks to have an optimized cost, in which nodes and edges are assigned a certain cost. In an embodiment of BLAST scoring, the reference matching probes ( Figure 2 The edges (marked by shading in the figure) ultimately have a relatively high cost. Since the edges may have a cost proportional to the length, the algorithm can seek to minimize the gap (as shown, the rectangular "probes" are nodes, and the arrows are edges). The nodes along the found path are the selected probe sets. In this algorithm, the selected probe sets provide non-overlapping coverage of the target RNA molecules.

[0043] Graph algorithms are known in the art, and are conducive to sequence analysis due to the natural fit between the linear nature of the genetic information carried in the nucleic acid sequence and the linear nature of the path through the directed graph. Graph platforms (such as Neo4J) are attractive here because they span the software / hardware barrier. Software packages can read, write and query (e.g., traverse) from the data stored as a graph, but graph storage makes the use of computer hardware non-standard. For example, some graph platforms use index-free adjacency, where the graph query reads the node, but the node contains a reference to the physical location in the physical memory at the adjacent node, so that the traversal moves to the reference position in the hardware. Compared with the relational database using the search index table written by the familiar software to record the relationship, this type of method using the spatial address on the hardware to perform index-free adjacency is much faster than the index table. Because the algorithm disclosed in the present invention is conducive to implementation in the graph platform, the method disclosed in the present invention is potentially very fast, and can be extended to query any reference target of any size, and FASTA probes can be performed BLAST for any reference group of any size without any difficulty.

[0044] In an exemplary embodiment, the cost function for traversing the graph is set to k*score^r (the k, r parameters can be specified from the command line, but by default k=256, r=2). In practice, when score=1 (the worst possible case), k is equal to the distance penalty that would be paid by leaving a gap of sqrt(k) bases. In the cost function, r defines the rate at which the score penalty decreases, so that larger values ​​of r mean more liberal handling of small differences / small mismatches. The graph traversal minimizes the cost function, and the nodes visited become the probe set. The probe associated with the graph traversal with the best score (not necessarily the absolute minimum) is used.

[0045] Figure 3 The distribution of inter-probe gap sizes for different k values ​​in the cost function is shown. Increasing k values ​​appear to reduce the overall BLAST score of the probes, which is a potential trade-off in terms of overall coverage. The probe sets can be manually inspected. For example, a biologist might choose k=1000 because any match to off-target RNA is scored very expensively, so the resulting probe set will not deplete rare mRNAs, even though it may leave a gap of 28 bases of undepleted ribosomal RNA after RNase H digestion (the final number for the k=1000 line in the figure).

[0046] Therefore, the software package first parses the target sequence to generate a set of probes (candidate sequences complementary to the RNA molecule within the predicted melting temperature boundary), performs BLAST on the probes to assign a cost function to each candidate sequence in the candidate sequence based at least in part on the match score with the reference off-target RNA, and finds a directed graph that minimizes the cost through the probes (candidate sequences) to thereby select a set of probes mapped to positions along the RNA molecule. As discussed, the group is selected by an algorithm that minimizes the cumulative cost function, gaps between positions, and overlaps. The described algorithm does not really minimize overlaps, but simply avoids the overlaps. Other algorithms are within the scope of the present disclosure. In fact, it is not necessarily necessary to avoid overlaps, because it is not assumed that the so-called overlapping probes actually compete to bind to the same copy of the molecule. Using method 101, a computer system selects probes to provide DNA oligonucleotides, each of which uniquely anneals to the RNA molecule, preferably with minimal or no overlap, and with minimal or no annealing to the off-target RNA.

[0047] The second probe design method BLAST filter is also within the scope of the present invention. In the embodiment of the BLAST filter, a computer system can be used to first use, for example, input target RNA sequence files and Perl scripts to generate all probes in a given Tm window. In any embodiment, the probe length can be set to have a default value (for example, 21), but the length can also vary, for example, different values ​​can be used, the value can be switched in operation or the range of the value can be set. Similarly, the temperature parameter can vary. For example, some embodiments use a script that allows 3.0 degrees of variance around the initial value (that is, if the minimum and maximum lengths do not allow the generation of probes with Tm initial values, the script will generate the next nearest candidate within 3.0 degrees---this can be adjusted or excluded). In order to generate all probes, the probe is designed from each base of 5' by finding the first probe of the desired Tm. The other strategy that can be used for any embodiment to ensure that there is no loss at the 3' end makes the script also produce probes from 3' end with 250 bases. In each preferred embodiment, the script makes the probe group generated initially unique.

[0048] In this BLAST filter embodiment, the system will BLAST the probes against the RefSeq-target sequence using blastn-short. Probes with percent identity above a given threshold are filtered out. The results show that this strategy gives results with some similarities to the Bloom filter method. These BLAST filtering methods can be used to generate probes with the most stringent possible identity criteria.

[0049] Any suitable method can be used to generate probe groups.An insight of the present invention is that first, information about the biochemical properties of target and off-target RNA is used to generate probe groups, which will actually perform the expected task of exhausting target RNA and leaving off-target RNA. After generating probes according to biophysical properties, practitioners can apply other criteria (e.g., the cost function of the Dijsktra algorithm that minimizes the gap), which promotes results similar to the main purpose (no gap) of the probe design strategy of the prior art. Any suitable method using biophysical properties can be used, such as the Bloom filter, BLAST filter or BLAST scoring method presented here. The method of the present invention provides probe groups that tend to have certain unique features.

[0050] Certain features of probe sets designed according to the methods of the present invention are unique and noteworthy. For example, the spacing between probes is highly variable and appears to be random (see Gap sequence below). Some probes abut one or more other probes (i.e., the 5' end touches the 3' end of another probe and / or the 3' end touches the 5' end of another probe). Some probes overlap with another probe (note comments on overlap elsewhere). Many gaps exceed 30 nucleotides (maximum = 111, see Figure 4 The probe length is variable (minimum = 18, maximum = 44, see Figure 5 ). The Tm of the probes was largely uniform, falling primarily into two discrete groups (it was generally found that probes targeting GC-rich segments needed to have higher Tm than probes targeting more balanced segments). Coverage of all targets was not end-to-end. In fact, coverage of 45S rRNA was missing 5 nucleotides at the 5' end and 47 nucleotides at the 3' end. The probes were not present in equimolar concentrations in the mixture - some concentrations were boosted up to 10-fold to improve performance (based on empirical data).

[0051] Some of the features are contrary to the generally accepted knowledge about exhaustion probes. For example, in the probe design of the prior art, gaps of 30 or even 111 bases are rare and are considered to be things to be avoided. Additionally, disordered mixing of gaps, adjacencies and overlaps is considered to be things to be avoided. It is noteworthy that the uniformity of length is sacrificed, and the probe of the present invention has a variable length and a substantially uniform melting temperature. Because the probe of the present invention is first designed according to biophysical properties, it is superior to the probe of the prior art (for example, substantially uniform Tm ensures the consistent digestion of RNase H, because the expected heteroduplex will all be stable under similar conditions, and in the strict tiling strategy of the prior art, the probe and target in the AT-rich zone may melt apart). One result of the design paradigm is that the probe spacing is highly variable and appears to be random.

[0052] The strategies disclosed herein have significant advantages. For example, there is a reduced need to denature the target and / or probe, thereby reducing workflow. The present invention eliminates the need to add enzymes when heating the reaction mixture, which may be due to more efficient denaturation of off-target probe binding. Overall, these reductions cut workflow time in half compared to competitors, with no significant loss in data quality.

[0053] Figure 4is a histogram of gap sizes determined for one specific RNA. As discussed, the probe spacing is highly variable and appears to be random. A set of probes is designed for a specific RNA.The following is a sequence of interprobe gap sizes where the probes form heteroduplexes with a particular RNA (negative values ​​indicate overlap): 7, 10, 10, 10, 1, 9, 29, 18, 14, 15, 3, 21, 0, 0, 0, 7, 12, 17, 2, 1, 2, 3, 2, 33, 6, 18, 16, 34, 6, 9, 11, 2, 19, 3, 20, 11, 1, 6, 18, 12, 10, 13, 11, 13, 17, 6, 6, 8, 16, 14, 19, 10, 0, 11, 8, 2, 14, 5, 7, 8, 3, 9, 11, 21, 0, 3, 17, 12, 13, 14, 15, 18 8,19,6,5,5,14,1,0,0,18,19,13,15,6,12,3,111,1,1,46,21,12,12,8,11,12,9,0,1,7,3,2,0,14,9,0,7,7,20,0,7,1,3,1,1,0,0,1,0,10,3,3, 26,4,11,9,10,1,2,1,5,0,0,5,19,7,17,10,5,6,0,3,8,5,0,6,6,6,7,13,5,29,0,21,29,3,7,5,12,7,8,12,0,9,4,7,9,5,34,3,13,10,9,9,4,22 ,36,15,26,29,8,14,17,4,44,12,11,47,9,10,9,28,25,15,0,19,22,15,37,14,12,33,5,29,39,45,7,41,25,13,32,33,10,16,39,0,0,3,18,24,7,20,7,4,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,25,0,0,0,0,0,0,0,0,-28,19,10,0,0,0,16,0,4,14,10,10,22,-9,0,18,4,0,0,3,8,2,4,14,8,2 ,2,6,3,5,8,8,25,25,14,4,5,-8,2,13,12,16,2,2,5,9,6,2,9,6,10,11,3,8,4,4,4,5,3,3,0,0,0,3,0,10,0,4,0,0,0,0,5,10,0,0,0,0,9,6,13 , 0, 1, -9, 8, 12, 13, 26, 11, 9, 45, -16, -5, -25, 4, 0, 18, 11, 0, 34, 14, 1, 0, 0, 0, 0, 1, 4, 8, 1, 0, 3, 5, 22, 14, 37, 11, 16, 9, 7, 0, 45, 10, 6, 16, 45, 22, 0, 28. There is no apparent uniformity or pattern.The probe sets were not designed to form patterns along the molecule.

[0054] Similar to the gap size, the probe length is variable.

[0055] Figure 5 A histogram of probe lengths for a set designed for a particular RNA is given (where d$Length is the probe length). The gap histogram and the probe length histogram reveal important properties of the probe set of the present invention. The gaps are not uniform. In fact, the distribution of gaps is not regular or standard.

[0056] From the histogram, it is noted that there is at least one gap exceeding 100. The mode may be zero. Moreover, there are negative values. Similarly, the probe length includes at least one at 44 and an obvious mode just exceeding 20. The histogram shows such a pattern because the probe is not designed to a gap or length specification. In fact, biophysical properties can not only inform sequence selection, but also inform probe length. For example, it is possible to allow the preparation of high GC oligonucleotides at the higher end within the Tm boundary, but it is also possible to prepare shorter than low GC oligonucleotides. Algorithmically, the system can seek to minimize the spacing between adjacent heteroduplexes, but still provide such a histogram (with a value of>100).

[0057] Using the probes disclosed herein, RNA depletion can be performed by hybridizing multiple DNA oligonucleotides to target RNA molecules in a sample to form heteroduplexes, wherein the DNA oligonucleotides have sequences selected to achieve the following: (i) the melting temperature of the heteroduplex is within a predetermined range, and (ii) the match with a reference off-target RNA is minimized; and digesting the RNA in the heteroduplex. These steps can be performed in a single-cell RNA-Seq (scRNA-Seq) workflow. For example, the method can include isolating single cells into fluid partitions, releasing nucleic acids from cells within the partitions (e.g., by chemical or thermal cell lysis), and depleting target RNA molecules. The fluid partitions can be, for example, droplets in an emulsion or microfluidic device, or, for example, wells in a plate (such as a multiwell plate or a pico titer plate). The method may include introducing reaction reagents (e.g., any of a cleavage reagent, a depletion probe of the invention, RNase H, a capture oligonucleotide, a primer, a template switching oligonucleotide, a tagging reagent, a reverse transcriptase, dNTPs, etc.) as the droplets form by pipetting into the wells or by droplet merging. Any of the primers or oligonucleotides may include a barcode, a sequencing adapter, a universal priming site, etc. For example, in some embodiments, each RNA transcript is captured by an oligonucleotide that contains a unique or nearly unique sequence to function as a unique molecular identifier, and each cDNA is given a partition barcode or a cell barcode.

[0058] Using the depletion probes of the present invention, heteroduplexes are formed and can show non-uniform distribution along the RNA molecules. RNase H digests the target RNA. Preferably, the DNA oligonucleotides (i.e., depletion probes) have been tested and shown to not stably hybridize with protein-coding transcripts in standardized human reference RNA extracts. After digestion, excess RNA (such as ribosomes or globin) is substantially removed. The remaining transcripts are copied into cDNA (optionally barcoded by molecules and / or cells). Downstream method steps can further copy or amplify cDNA molecules. For example, cDNA molecules can be amplified to produce sequencing libraries, for example, using sequencing adapters such as llumina Y adapters at the ends of library members.

[0059] The methods described herein can be used to provide a set of RNA depletion probes, the set of RNA depletion probes comprising a plurality of DNA oligonucleotides capable of hybridizing with a target RNA molecule to form a heteroduplex. The DNA oligonucleotides have sequences selected to achieve the following: (i) the melting temperature of the heteroduplex is within a predetermined range, and (ii) the matching with the off-target RNA is minimized. Preferably, the heteroduplex will be formed along the RNA molecule with a non-uniform distribution, for example, the spacing between the heteroduplexes can be highly variable and / or mathematically random. The probes can be designed by a computer system, even automatically, by an algorithm, the computer system being operable to generate a set of candidate sequences complementary to the RNA molecule, wherein the predicted melting temperature is within a predetermined range; assigning a cost function to each candidate sequence in the candidate sequence based at least in part on a matching score with a reference off-target RNA; and selecting a set of candidate sequences mapped to positions along the RNA molecule, wherein the group is selected by an algorithm that minimizes the cumulative cost function, gaps between positions, and overlaps. The computer system can export the selected group as any suitable file, such as FASTA file, formatting for inputting into the file in the oligonucleotide synthesis instrument, or with the format accepted by commercial oligonucleotide supplier (such as integrated DNA technology company (Integrated DNA Technologies, IDT)).Even in some embodiments, the computer system can also include an application programming interface, which selects probe groups and transmits well-formed IDT commands by internet protocol.Once synthesized, probe groups can be used to provide in packaging, which is used in the laboratory or used for transportation or sale.For example, probe groups can be freeze-dried in a solution in a pipe or in a pipe, and wherein the pipe can be any suitable pipe, such as 1.5mL microcentrifuge tubes (such as the pipe sold under the trademark EPPENDORF).Then, probe groups can be used for the method for RNA depletion as described above.

[0060] As mentioned above, probe groups can be selected by using a Bloom filter, a BLAST filter or a BLAST scoring method. Preferred BLAST scoring embodiments are described in detail above. Bloom filter embodiments are described in the examples. In addition to Bloom filters, BLAST filters or BLAST scoring methods or any other suitable methods can also be used to provide depletion probes. The present invention provides RNA depletion probes, i.e., groups of short DNA oligonucleotides, which hybridize along the length of the target RNA and mediate the digestion of the target RNA by RNase H to remove excess RNA molecules from a sample. Depletion probes according to the present invention are designed mainly based on the biochemical and biophysical properties of the probes, so that all depletion probes in a group of depletion probes show substantially uniform and consistent behavior in terms of binding to the target RNA in the sample. Probes are designed primarily for specific performance targets and biophysical properties, thereby producing probe groups with irregular, even substantially random spacing along the target RNA molecule.

[0061] Example

[0062] Example 1. Bloom filter

[0063] Figure 6 The workflow for the bloom filter design method is illustrated. The workflow begins with setting up the target RNA and having a RefSeq data set. For example, a package such as BioPerl can be used to open a handle to the latest version of RefSeq, as described in Pruitt, 2007, NCBI reference sequences (RefSeq): a curated non-redundants sequence database of genomes, transcripts and proteins, Nucleic Acids Res 35(Database): D61-5, which is incorporated by reference. The target RNA is stripped from (or masked in) RefSeq.

[0064] Additionally, the target is k-aggregated, or parsed into k-aggregates, such as all possible k-aggregates. It is easy for a technician to write a script (such as a Perl script) that writes all possible k-aggregates into a FASTA file. Depending on the platform, it may be preferred that the FASTA file is converted into a binary alignment map (BAM) for downstream processing, such as by the PathSeq pipeline.

[0065] Using a bloom filter, search for k-mers from the RefSeq in the target. This can be done using the PathSeq GATK pipeline to build a bloom filter package that identifies k-mers present in the target sequence that are not present elsewhere in the RefSeq.

[0066] A Bloom filter is a space-efficient probabilistic data structure that can be used to test whether an element is a member of a group. False positive matches are possible, but false negatives are impossible. An empty Bloom filter is an m-bit bit array with all bits set to 0. It is also necessary to define k different hash functions, each of which maps or hashes a set of elements to one of the m array positions, thereby generating a uniform random distribution. Typically, k is a small constant, which depends on the desired false error rate ε, and m is proportional to k and the number of elements to be added. To add an element, it is fed to each of the k hash functions to obtain k array positions. The bits at all these positions are set to 1. To query an element (test whether it is in the group), it is fed to each of the k hash functions to obtain k array positions. If any of the bits at these positions is 0, the element is definitely not in the group; if it is in the group, all bits will be set to 1 when inserted. If all are 1, the element is in the group, or the bit is accidentally set to 1 during the insertion of other elements, resulting in a false positive. In a simple Bloom filter, it is impossible to distinguish between these two cases, but more advanced techniques can solve this problem. See Najam, 2019, Pattern matching for DNA sequencing data using multiple Bloom filters, Biomed Res Int 14:7074387, which is incorporated by reference.

[0067] Bloom filters are applied to k-mers. For example, a Bloom filter can be used to simply exclude k-mers found in RefSeq. Alternatively, a Bloom filter can be used to find k-mers in RefSeq and assign scores to those k-mers (e.g., some constant multiplied by the match % or some other such arbitrary score).

[0068] Through the k-polymer group in the file or data structure, the system tiles a group of k-polymers across the target RNA. This can be performed as a graph traversal, and due to the unique way graph platform (e.g., with index-free adjacency) such as Neo4J access hardware, the graph traversal is very suitable for being executed by a computer, even for large (including very very large) data volumes. The system tiles RNA sequences. Tiling (RNA) sequences can be stated as a problem of traversing a directed graph in which the edge has a cost. In certain embodiments, the cost of the initial edge is the distance (in bases) between the probe end and the next probe start. Using Dijkstra's algorithm, the lowest cost path from RNA to the end can be calculated most efficiently (and optimally) (see, for example, Rodriguez-Puente, 2013, for performing the shortest path search in geographic information systems by using reduced graphs, Springerplus 2: 291, the document is incorporated by reference). Dijkstra runs in O[|V|+|E|*log(|V|)]. Using a simple heuristic that only considers probes that exist within a fixed distance of each other (rather than all probes vs. all probes), |V|∈O(|E|), so the running time complexity is O[|E|*log(|E|)].

[0069] The Dijkstra algorithm is implemented in a general manner that allows it to operate on any graph object, for which nodes can hold any data type. It is noteworthy that in a first embodiment of the tiling algorithm, a graph is constructed so that the distance from probe x to probe y is start (y) - stop (x). This is feasible, but Dijkstra is a greedy algorithm and may leave gaps at the end of the tiling path (i.e., when there are multiple optimal solutions, it will select a solution in which costs are generated in the later stages of the optimization, and these costs will usually be accumulated in longer segments). This can be solved relatively easily by setting the distance between nodes to [start (y) - stop (x)]^2, so that when a single gap is extended relative to opening multiple short gaps, the cost grows faster. Additionally or alternatively, it may be appropriate to assign a gap extension penalty that is different from the gap formation penalty.

[0070] It is worth noting that the graph algorithm can read from the initial probe candidates that cover each base in the target (e.g., RNA18SN3) while leaving uncovered long segments. For RNA18SN3, all filtered 21-mers cover all bases except 4 bases, but two Dijkstra tiles leave 223 and 222 uncovered bases respectively. Since the probes are non-overlapping and only one probe starts from each position, the probes are filtered out to make it almost impossible to find a continuous coverage path. Interestingly, this is not necessarily a limitation to applicability, and the resulting probe set can be used for RNA depletion. It is worth noting that in the existing algorithm, when an unfiltered probe set is used (i.e., when each possible probe is fed into the tiler), the return is the best path with only a small coverage gap at the end (if the length of the tiled sequence is not a multiple of the probe length, the length of the probe is all fixed in this case). The input containing the correct tile returns the same input. Manual management results may be helpful, but manual management results seem to confirm that suitable results have been obtained.

[0071] Example 2. Design workflow

[0072] Figure 7 Examples of human / mouse / rat (H / M / R) probe design methods are described. The methods include: algorithmic design of depletion probes for human transcript targets; removal of probes with high homology to undesirable human transcript targets; algorithmic design of depletion probes for mouse / rat transcript targets; removal of probes with high homology to undesirable mouse / rat transcript targets; cross-checking human probes with mouse / rat transcriptomes and vice versa; and removal of probes with high homology to undesirable transcript targets in H / M / R. After such design methods, the final probe set can be synthesized and used to deplete unwanted RNA from a sample, or packaged for storage, transportation, distribution or subsequent use. For example, the probe can be placed in a tube or vial, in an aqueous solution or lyophilized for subsequent reconstruction. Such vials or tubes can optionally include other reagents, such as RNase H.

[0073] Example 3. Results

[0074] Figure 8 Depletion of ribosomal RNA in total RNA from whole blood is shown. The left bar "Before RNA depletion" shows that the sample has >80% ribosomal RNA. After rRNA depletion, as shown in the right bar, the sample has <5% rRNA and about 75% mRNA. Therefore, these results show that the probe sets and methods of the present disclosure are effective for removing ribosomal RNA from a sample.

[0075] Fig. 9Results of depletion of hemoglobin RNA from whole blood using the probe sets and methods of the present disclosure are shown. Prior to hemoglobin depletion, as shown in the left bar, the sample contained approximately 35.75% residual hemoglobin RNA. After depletion, the sample contained approximately 0.01% residual hemoglobin RNA. Thus, these results show that the probe sets and methods of the present disclosure are effective for removing hemoglobin RNA from a sample.

[0076] Fig.10 It was shown that the methods and probe sets of the invention can be used to robustly deplete ribosomal RNA from total RNA samples derived from formalin-fixed paraffin-embedded (FFPE) tissue with 10 ng of RNA.

[0077] Fig.11 Results for FFPE with 400 ng RNA are shown.

[0078] For each of the 4 blocks, 10 ng and 400 ng of total RNA were obtained per block, and in all cases, the ribosomal RNA% after depletion was less than or equal to about 7%, and in most cases, significantly less than 5%. Therefore, these results show that the probe sets and methods of the present disclosure are effective for removing RNA from FFPE samples.

[0079] Fig.12 Quantitative results for ribosomal depletion in total RNA derived from a universal mouse and rat reference are presented. Using the methods and probe sets of the present invention, the mouse sample was depleted from about 85% ribosomal RNA down to about less than %% ribosomal RNA. The rate sample had an original about 80% ribosomal RNA, and the methods and probe sets of the present invention depleted the rat sample to less than about 5% ribosomal RNA. Therefore, these results show that the probe sets and methods of the present disclosure are effective for removing RNA from rat or mouse samples.

Claims

1. A method for depleting RNA, the method comprising: exposing a plurality of DNA oligonucleotides to a sample, wherein the DNA oligonucleotides form heteroduplexes with a specific subset of RNAs in the sample within a predetermined melting temperature range, but do not substantially interact with RNAs that are not part of the subset; and The RNA in the resulting heteroduplex is digested.

2. The method of claim 1, wherein a plurality of said DNA oligonucleotides form heteroduplexes along the length of the RNA molecules in said subset.

3. The method of claim 2, wherein the DNA oligonucleotides form heteroduplexes in a non-uniform distribution along the RNA molecules in the subset.

4. The method of claim 1, wherein the DNA oligonucleotides form heteroduplexes that are tiled along the length of the members of the RNA subset.

5. The method of claim 4, wherein the oligonucleotides are designed to minimize the spacing between adjacent heteroduplexes.

6. The method of claim 1, wherein the predetermined melting temperature range is optimized for a specific sequence of the DNA oligonucleotide to maximize binding to the RNA in the subset.

7. The method of claim 1, wherein the sequence of the DNA oligonucleotide is selected by: generating a set of candidate sequences complementary to the RNA in the subset, wherein the melting temperature is within the predetermined range; assigning a cost function to each of the candidate sequences based at least in part on a match score to a reference off-target RNA; and A set of the candidate sequences is selected that minimizes the cost function, inter-position gaps, and overlaps.

8. The method of claim 1, wherein the RNA subset is selected from the group consisting of ribosomal RNA, globin transcripts, and mitochondrial RNA.

9. The method of claim 1, wherein the digestion step comprises treating the sample with RNAseH.

10. The method of claim 1, wherein the method increases the ratio of poly-A tail RNA to non-coding RNA in the sample.

11. The method of claim 1, wherein the spacing between the heteroduplexes is mathematically random.

12. The method of claim 1, wherein the sequences of at least two of the DNA oligonucleotides are selected to hybridize to the RNA subset at overlapping or adjacent positions.

13. The method of claim 1, wherein the DNA oligonucleotide has a length ranging from about 18 bases to about 44 bases.

14. The method of claim 1, wherein the DNA oligonucleotides are present in the mixture at non-equimolar concentrations.

15. The method according to claim 14, wherein the RNA, the DNA oligonucleotide and the RNaseH are added together into a single tube before performing the hybridization and digestion.

Images