SNP Prediction Research Articles

Forensic DNA Phenotyping: a review on SNP Panels, Genotyping Techniques, and Prediction Models

Abstract In the past few years, forensic DNA phenotyping has attracted a strong interest in the forensic research. Among the increasing publications, many have focused on testing the available panels to infer biogeographical ancestry on less represented populations and understanding the genetic mechanisms underlying externally visible characteristics. However, there are currently no publications that gather all the existing panels limited to forensic DNA phenotyping and discuss the main technical limitations of the technique. In this review, we performed a bibliographic search in Scopus database of phenotyping-related literature, which resulted in a total of 48, 43 and 15 panels for biogeographical ancestry, externally visible characteristics and both traits inference, respectively. Here we provide a list of commercial and non-commercial panels and the limitations regarding the lack of harmonization in terms of terminology (i.e., categorization and measurement of traits) and reporting, the lack of genetic knowledge and environment influence to select markers and develop panels, and the debate surrounding the selection of genotyping technologies and prediction models and algorithms. In conclusion, this review aims to be an updated guide and to present an overview of the current related literature.

In silico SNP prediction of selected protein orthologues in insect models for Alzheimer's, Parkinson's, and Huntington’s diseases

Alzheimer's, Parkinson’s, and Huntington’s are the most common neurodegenerative diseases that are incurable and affect the elderly population. Discovery of effective treatments for these diseases is often difficult, expensive, and serendipitous. Previous comparative studies on different model organisms have revealed that most animals share similar cellular and molecular characteristics. The meta-SNP tool includes four different integrated tools (SIFT, PANTHER, SNAP, and PhD-SNP) was used to identify non synonymous single nucleotide polymorphism (nsSNPs). Prediction of nsSNPs was conducted on three representative proteins for Alzheimer's, Parkinson’s, and Huntington’s diseases; APPl in Drosophila melanogaster, LRRK1 in Aedes aegypti, and VCPl in Tribolium castaneum. With the possibility of using insect models to investigate neurodegenerative diseases. We conclude from the protein comparative analysis between different insect models and nsSNP analyses that D. melanogaster is the best model for Alzheimer’s representing five nsSNPs of the 21 suggested mutations in the APPl protein. Aedes aegypti is the best model for Parkinson’s representing three nsSNPs in the LRRK1 protein. Tribolium castaneum is the best model for Huntington’s disease representing 13 SNPs of 37 suggested mutations in the VCPl protein. This study aimed to improve human neural health by identifying the best insect to model Alzheimer's, Parkinson’s, and Huntington’s.

Mining Salt Tolerance SNP Loci and Prediction of Candidate Genes in the Rice Bud Stage by Genome-Wide Association Analysis.

Mining salt tolerance genes is significant for breeding high-quality salt-tolerant rice varieties in order to improve the utilization of saline-alkaline land. In this study, 173 rice accessions were measured for their germination potential (GP), germination rate (GR), seedling length (SL), root length (RL), germination potential relative to salt damage rate (GPR), germination rate relative to salt damage rate (GRR), seedling length relative to salt damage rate (SLR), relative salt damage rate at the germination stage (RSD) and comprehensive relative salt damage rate in the early seedling stage (CRS) under normal and salt stress conditions. Genome-wide association analysis was performed with 1,322,884 high-quality SNPs obtained by resequencing. Eight quantitative trait loci (QTLs) related to salt tolerance traits at the germination stage were detected in 2020 and 2021. They were related to the GPR (qGPR2) and SLR (qSLR9), which were newly discovered in this study. Three genes were predicted as salt tolerance candidate genes: LOC_Os02g40664, LOC_Os02g40810, and LOC_Os09g28310. At present, marker-assisted selection (MAS) and gene-edited breeding are becoming more widespread. Our discovery of candidate genes provides a reference for research in this field. The elite alleles identified in this study may provide a molecular basis for cultivating salt-tolerant rice varieties.

In silico analysis predicting effects of deleterious SNPs of human RASSF5 gene on its structure and functions

Ras association domain-containing protein 5 (RASSF5), one of the prospective biomarkers for tumors, generally plays a crucial role as a tumor suppressor. As deleterious effects can result from functional differences through SNPs, we sought to analyze the most deleterious SNPs of RASSF5 as well as predict the structural changes associated with the mutants that hamper the normal protein–protein interactions. We adopted both sequence and structure based approaches to analyze the SNPs of RASSF5 protein. We also analyzed the putative post translational modification sites as well as the altered protein–protein interactions that encompass various cascades of signals. Out of all the SNPs obtained from the NCBI database, only 25 were considered as highly deleterious by six in silico SNP prediction tools. Among them, upon analyzing the effect of these nsSNPs on the stability of the protein, we found 17 SNPs that decrease the stability. Significant deviation in the energy minimization score was observed in P350R, F321L, and R277W. Besides this, docking analysis confirmed that P350R, A319V, F321L, and R277W reduce the binding affinity of the protein with H-Ras, where P350R shows the most remarkable deviation. Protein–protein interaction analysis revealed that RASSF5 acts as a hub connecting two clusters consisting of 18 proteins and alteration in the RASSF5 may lead to disassociation of several signal cascades. Thus, based on these analyses, our study suggests that the reported functional SNPs may serve as potential targets for different proteomic studies, diagnosis and therapeutic interventions.

MethylToSNP: identifying SNPs in Illumina DNA methylation array data

BackgroundCurrent array-based methods for the measurement of DNA methylation rely on the process of sodium bisulfite conversion to differentiate between methylated and unmethylated cytosine bases in DNA. In the absence of genotype data this process can lead to ambiguity in data interpretation when a sample has polymorphisms at a methylation probe site. A common way to minimize this problem is to exclude such potentially problematic sites, with some methods removing as much as 60% of array probes from consideration before data analysis.ResultsHere, we present an algorithm implemented in an R Bioconductor package, MethylToSNP, which detects a characteristic data pattern to infer sites likely to be confounded by polymorphisms. Additionally, the tool provides a stringent reliability score to allow thresholding on SNP predictions. We calibrated parameters and thresholds used by the algorithm on simulated and real methylation data sets. We illustrate findings using methylation data from YRI (Yoruba in Ibadan, Nigeria), CEPH (European descent) and KhoeSan (southern African) populations. Our polymorphism predictions made using MethylToSNP have been validated through SNP databases and bisulfite and genomic sequencing.ConclusionsThe benefits of this method are threefold. First, it prevents extensive data loss by considering only SNPs specific to the individuals in the study. Second, it offers the possibility to identify new polymorphisms in samples for which there is little known about the genetic landscape. Third, it identifies variants as they exist in functional regions of a genome, such as in CTCF (transcriptional repressor) sites and enhancers, that may be common alleles or personal mutations with potential to deleteriously affect genomic regulatory activities. We demonstrate that MethylToSNP is applicable to the Illumina 450K and Illumina 850K EPIC array data and is also backwards compatible to the 27K methylation arrays. Going forward, this kind of nuanced approach can increase the amount of information derived from precious data sets by considering samples of the project individually to enable more informed decisions about data cleaning.

A high-throughput SNP discovery strategy for RNA-seq data

BackgroundSingle nucleotide polymorphisms (SNP) have been applied as important molecular markers in genetics and breeding studies. The rapid advance of next generation sequencing (NGS) provides a high-throughput means of SNP discovery. However, SNP development is limited by the availability of reliable SNP discovery methods. Especially, the optimum assembler and SNP caller for accurate SNP prediction from next generation sequencing data are not known.ResultsHerein we performed SNP prediction based on RNA-seq data of peach and mandarin peel tissue under a comprehensive comparison of two paired-end read lengths (125 bp and 150 bp), five assemblers (Trinity, IDBA, oases, SOAPdenovo, Trans-abyss) and two SNP callers (GATK and GBS). The predicted SNPs were compared with the authentic SNPs identified via PCR amplification followed by gene cloning and sequencing procedures. A total of 40 and 240 authentic SNPs were presented in five anthocyanin biosynthesis related genes in peach and in nine carotenogenic genes in mandarin. Putative SNPs predicted from the same RNA-seq data with different strategies led to quite divergent results. The rate of false positive SNPs was significantly lower when the paired-end read length was 150 bp compared with 125 bp. Trinity was superior to the other four assemblers and GATK was substantially superior to GBS due to a low rate of missing authentic SNPs. The combination of assembler Trinity, SNP caller GATK, and the paired-end read length 150 bp had the best performance in SNP discovery with 100% accuracy both in peach and in mandarin cases. This strategy was applied to the characterization of SNPs in peach and mandarin transcriptomes.ConclusionsThrough comparison of authentic SNPs obtained by PCR cloning strategy and putative SNPs predicted from different combinations of five assemblers, two SNP callers, and two paired-end read lengths, we provided a reliable and efficient strategy, Trinity-GATK with 150 bp paired-end read length, for SNP discovery from RNA-seq data. This strategy discovered SNP at 100% accuracy in peach and mandarin cases and might be applicable to a wide range of plants and other organisms.

In silico approaches to discover the functional impact of non-synonymous single nucleotide polymorphisms in selective sweep regions of the Landrace genome.

ObjectiveThe aim of this study was to discover the functional impact of non-synonymous single nucleotide polymorphisms (nsSNPs) that were found in selective sweep regions of the Landrace genomeMethodsWhole-genome re-sequencing data were obtained from 40 pigs, including 14 Landrace, 16 Yorkshire, and 10 wild boars, which were generated with the Illumina HiSeq 2000 platform. The nsSNPs in the selective sweep regions of the Landrace genome were identified, and the impacts of these variations on protein function were predicted to reveal their potential association with traits of the Landrace breed, such as reproductive capacity.ResultsTotal of 53,998 nsSNPs in the mapped regions of pigs were identified, and among them, 345 nsSNPs were found in the selective sweep regions of the Landrace genome which were reported previously. The genes featuring these nsSNPs fell into various functional categories, such as reproductive capacity or growth and development during the perinatal period. The impacts of amino acid sequence changes by nsSNPs on protein function were predicted using two in silico SNP prediction algorithms, i.e., sorting intolerant from tolerant and polymorphism phenotyping v2, to reveal their potential roles in biological processes that might be associated with the reproductive capacity of the Landrace breed.ConclusionThe findings elucidated the domestication history of the Landrace breed and illustrated how Landrace domestication led to patterns of genetic variation related to superior reproductive capacity. Our novel findings will help understand the process of Landrace domestication at the genome level and provide SNPs that are informative for breeding.

Accuracy of SNPs to predict risk of HLA alleles associated with drug-induced hypersensitivity events across racial groups.

To evaluate the potential usefulness of selected SNPs to predict specific HLA alleles that are associated with drug-induced hypersensitivity reactions (HSR) in different ethnic groups. Five specific HLA alleles known to predict HSR were tagged by seven SNPs (rs1061235-HLA-A*31:01; rs2395029-HLA-B*57:01; rs3909184-HLA-B*15:02; rs9469003-HLA-B*58:01; rs3117583-HLA-B*58:01; rs9270986-HLA-DQA1*01:02 and rs3129900-HLA-DQA1*01:02). DNA from 24 African-Americans, 56 Asian, 44 Caucasians and 36 Hispanics of known high resolution HLA-A, B and DQA1 status were genotyped for tagSNPs using TaqMan. Sensitivity and specificity were considered the primary end points and were 100% across the four populations for rs2395029-HLA-B*57:01. SNP prediction of HLA-A*31:01 had 100% sensitivity and 84% specificity. This study demonstrates the utility of SNP tagging as a 'real time' approach to predict or exclude the presence of specific HLA alleles of known importance to HSR across diverse ethnic groups. Original submitted 24 April 2014; Revision submitted 2 April 2015.

Use of long term molecular dynamics simulation in predicting cancer associated SNPs.

Computational prediction of cancer associated SNPs from the large pool of SNP dataset is now being used as a tool for detecting the probable oncogenes, which are further examined in the wet lab experiments. The lack in prediction accuracy has been a major hurdle in relying on the computational results obtained by implementing multiple tools, platforms and algorithms for cancer associated SNP prediction. Our result obtained from the initial computational compilations suggests the strong chance of Aurora-A G325W mutation (rs11539196) to cause hepatocellular carcinoma. The implementation of molecular dynamics simulation (MDS) approaches has significantly aided in raising the prediction accuracy of these results, but measuring the difference in the convergence time of mutant protein structures has been a challenging task while setting the simulation timescale. The convergence time of most of the protein structures may vary from 10 ns to 100 ns or more, depending upon its size. Thus, in this work we have implemented 200 ns of MDS to aid the final results obtained from computational SNP prediction technique. The MDS results have significantly explained the atomic alteration related with the mutant protein and are useful in elaborating the change in structural conformations coupled with the computationally predicted cancer associated mutation. With further advancements in the computational techniques, it will become much easier to predict such mutations with higher accuracy level.

Detection of Damaging nsSNPs on Human Caspase-Cascades Related to Apoptotic Signalling Pathway

In tumorigenesis, cancer genetics and the related mutations have been the main topic of study these days. Caspases have been found to be actively involved in the process of apoptosis. Malfunction of apoptosis is one of the causes for cancerous tumors and different caspase mutations are related to that process. It has been found that two groups of caspases involved in this process apoptosis which are initiator caspases and executioner caspases. SNPs have been extensively studied over the last decade, due to their association with a number of genetic diseases. Human SNPs have always been a source of information related to the complex changes associated with their origin. SNPs which can change the resulting amino acid i.e., nonsynonymous SNPs (nsSNPs) are of prime concern these days because of their direct relation with the disease or the respective individual. In this study our focus is not only to detect the nsSNPs available in the human caspase data but to further evaluate the potentially damaging nsSNPs. Using the computational approach we have been able to obtain almost seventy eight nsSNPs, among these few of the nsSNPs seem to have serious consequences, as they have been cross verified from a variety of SNP prediction tools. The functional as well as structural impact of the nsSNPs is determined and discussed. Our predicted nsSNPs on human caspases may be associated with cancer risk.

In silico prediction of a disease-associated STIL mutant and its affect on the recruitment of centromere protein J (CENPJ)

Human STIL (SCL/TAL1 interrupting locus) protein maintains centriole stability and spindle pole localisation. It helps in recruitment of CENPJ (Centromere protein J)/CPAP (centrosomal P4.1-associated protein) and other centrosomal proteins. Mutations in STIL protein are reported in several disorders, especially in deregulation of cell cycle cascades. In this work, we examined the non-synonymous single nucleotide polymorphisms (nsSNPs) reported in STIL protein for their disease association. Different SNP prediction tools were used to predict disease-associated nsSNPs. Our evaluation technique predicted rs147744459 (R242C) as a highly deleterious disease-associated nsSNP and its interaction behaviour with CENPJ protein. Molecular modelling, docking and molecular dynamics simulation were conducted to examine the structural consequences of the predicted disease-associated mutation. By molecular dynamic simulation we observed structural consequences of R242C mutation which affects interaction of STIL and CENPJ functional domains. The result obtained in this study will provide a biophysical insight into future investigations of pathological nsSNPs using a computational platform.

SalmonDB: a bioinformatics resource for Salmo salar and Oncorhynchus mykiss

SalmonDB is a new multiorganism database containing EST sequences from Salmo salar, Oncorhynchus mykiss and the whole genome sequence of Danio rerio, Gasterosteus aculeatus, Tetraodon nigroviridis, Oryzias latipes and Takifugu rubripes, built with core components from GMOD project, GOPArc system and the BioMart project. The information provided by this resource includes Gene Ontology terms, metabolic pathways, SNP prediction, CDS prediction, orthologs prediction, several precalculated BLAST searches and domains. It also provides a BLAST server for matching user-provided sequences to any of the databases and an advanced query tool (BioMart) that allows easy browsing of EST databases with user-defined criteria. These tools make SalmonDB database a valuable resource for researchers searching for transcripts and genomic information regarding S. salar and other salmonid species. The database is expected to grow in the near feature, particularly with the S. salar genome sequencing project.Database URL: http://genomicasalmones.dim.uchile.cl/

A complex crosstalk between polymorphic microRNA target sites and AD prognosis

Genetic variation associated with different diseases interferes with microRNA-mediated regulation by creating, destroying, or modifying microRNA (miRNA) binding sites. Alzheimer’s disease (AD) is a common neurodegenerative disorder that affects cognitive function in the elderly and has associated genetic variation within the diseased genome. We investigated the extent to which these SNPs interfere with miRNA gene regulation and affect AD susceptibility. We designed a sophisticated computational pipeline to predict the potential miRNA-target SNP interactions followed by a rigorous bioinformatic analysis to establish the effect of “gain-of-function” or “loss-of-function” of the miRNA mediated regulatory network on the etiology of the disease. The significant interactions include target SNPs present in seven genes-related to AD prognosis with the miRNAs- miR-214, -23a & -23b, -486-3p, -30e*, -143, -128, -27a &-27b, -324-5p and -422a. Further our study provides novel insight on the contribution of synonymous mutations towards AD pathogenesis. The putative miRNA::target SNP interactions suggest that the dysregulated miRNA network contributes to the aberrant gene expression in AD. Our results show that miRNA-target variability is a ubiquitous phenomenon in the adult human brain, which may influence gene expression in physiological and pathological conditions. The experimental assay of these miRNA::target SNP predictions will be the next step towards determining the functional effects of these SNPs on miRNA targets and their association with disease susceptibility. Overall, this work may serve as the cornerstones for understanding the variables of the molecular pathways relevant to the development of the disease and designing therapeutic miRNA regimens to modify diseased gene expression profiles in AD patients.

Abstract LB-264: Whole genome sequencing of an adrenocortical carcinoma reveals a pathogenic context and exposes therapeutic options

Abstract Adrenocortical carcinoma (ACC) is an aggressive endocrine malignancy with a 5-year survival rate of 20–30%. There are few treatment options for patients in part because the rarity of the tumor makes its study difficult. Previous work in our lab and by others has highlighted the importance of IGF2-IGF1R signaling, beta-catenin signaling, dysregulation of p53, and the G2/M transition in ACC pathogenesis. In order to advance knowledge of ACC biology and facilitate drug target discovery, we performed whole genome sequencing on an ACC tumor and matched peripheral blood. An ACC sample was obtained during surgery on a 51 year old female who had locally recurrent disease that progressed despite chemotherapy. DNA extracted from both tumor and blood was processed independently, and libraries were paired-end sequenced on the Illumina HiSeq 2000 platform. Sequencing data was aligned to the reference genome, build hg18. Variations were predicted in the tumor and blood separately using two different SNP calling algorithms: samtools and SolSNP. After SNP prediction, results were combined to identify variations called by both algorithms, submitted to PolyPhen2 to assess predicted functional consequence, and filtered to yield those SNPs most likely to represent deleterious, non-synonymous changes in coding regions of the genome in the tumor specifically, as well as those shared by the tumor and blood (i.e. constitutional changes).Our preliminary analysis identified a total of 1,724,133 single nucleotide changes in peripheral blood and 2,820,476 in the tumor prior to filtering. The transition to transversion ratio was 2.2 for blood and 2.0 for tumor. We did not observe mutations in known tumor suppressor or oncogenes. There were no coding mutations observed in any of the genes in the canonical pathways for IGF2 or beta-catenin signaling; p53 was wild-type. Inspection of constitutional changes indicates that the patient harbors a constitutive putative mutation in E2F2. In mice, knock-outs of E2F transcription factors result in developmental problems within the adrenal cortex. In our preliminary analysis, we identified a network that may possibly explain the aberrant p53 function that characterizes ACC. The tumor harbors a predicted mutation in SETD8. SETD8 encodes Set8, a lysine methyltransferase that has been implicated in gene silencing via histone 4 methylation, and is a negative regulator of p53 through methylation of p53 at lysine 382. Loss of p53 activity leads to increased activity of c-myc and SP1, both of which drive the expression of E2F2. Further work is required to characterize whether and how the mutations in SETD8 and E2F2 alter protein function and play a role in ACC pathogenesis. Finally, putative mutations in genes involved homologous repair such as ATM, FANCM, and BRCA2 suggests DNA cross-linking agents like mitomycin C may be useful in treating this patient's cancer. Data analysis is ongoing. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr LB-264. doi:10.1158/1538-7445.AM2011-LB-264

Repeat subtraction-mediated sequence capture from a complex genome

Sequence capture technologies, pioneered in mammalian genomes, enable the resequencing of targeted genomic regions. Most capture protocols require blocking DNA, the production of which in large quantities can prove challenging. A blocker-free, two-stage capture protocol was developed using NimbleGen arrays. The first capture depletes the library of repetitive sequences, while the second enriches for target loci. This strategy was used to resequence non-repetitive portions of an approximately 2.2 Mb chromosomal interval and a set of 43 genes dispersed in the 2.3 Gb maize genome. This approach achieved approximately 1800-3000-fold enrichment and 80-98% coverage of targeted bases. More than 2500 SNPs were identified in target genes. Low rates of false-positive SNP predictions were obtained, even in the presence of captured paralogous sequences. Importantly, it was possible to recover novel sequences from non-reference alleles. The ability to design novel repeat-subtraction and target capture arrays makes this technology accessible in any species.

Estimating Effects and Making Predictions from Genome-Wide Marker Data

In genome-wide association studies (GWAS), hundreds of thousands of genetic markers (SNPs) are tested for association with a trait or phenotype. Reported effects tend to be larger in magnitude than the true effects of these markers, the so-called ``winner's curse.'' We argue that the classical definition of unbiasedness is not useful in this context and propose to use a different definition of unbiasedness that is a property of the estimator we advocate. We suggest an integrated approach to the estimation of the SNP effects and to the prediction of trait values, treating SNP effects as random instead of fixed effects. Statistical methods traditionally used in the prediction of trait values in the genetics of livestock, which predates the availability of SNP data, can be applied to analysis of GWAS, giving better estimates of the SNP effects and predictions of phenotypic and genetic values in individuals.

Combining the interactome and deleterious SNP predictions to improve disease gene identification

A method has been developed for the prediction of proteins involved in genetic disorders. This involved combining deleterious SNP prediction with a system based on protein interactions and phenotype distances; this is the first time that deleterious SNP prediction has been used to make predictions across linkage-intervals. At each step we tested and selected the best procedure, revealing that the computationally expensive method of assigning medical meta-terms to create a phenotype distance matrix was outperformed by a simple word counting technique. We carried out in-depth benchmarking with increasingly stringent data sets, reaching precision values of up to 75% (19% recall) for 10-Mb linkage-intervals (averaging 100 genes). For the most stringent (worst-case) data we attained an overall recall of 6%, yet still achieved precision values of up to 90% (4% recall). At all levels of stringency and precision the addition of predicted deleterious SNPs was shown to increase recall.

Slider—maximum use of probability information for alignment of short sequence reads and SNP detection

Motivation: A plethora of alignment tools have been created that are designed to best fit different types of alignment conditions. While some of these are made for aligning Illumina Sequence Analyzer reads, none of these are fully utilizing its probability (prb) output. In this article, we will introduce a new alignment approach (Slider) that reduces the alignment problem space by utilizing each read base's probabilities given in the prb files.Results: Compared with other aligners, Slider has higher alignment accuracy and efficiency. In addition, given that Slider matches bases with probabilities other than the most probable, it significantly reduces the percentage of base mismatches. The result is that its SNP predictions are more accurate than other SNP prediction approaches used today that start from the most probable sequence, including those using base quality.Contact: nmalhis@bcgsc.caSupplementary information and availability: http://www.bcgsc.ca/platform/bioinfo/software/slider

Confirming Single Nucleotide Polymorphisms from Expressed Sequence Tag Datasets Derived from Three Cattle cDNA Libraries

Using the Phred/Phrap/Polyphred/Consed pipeline established in the National Livestock Research Institute of Korea, we predicted candidate coding single nucleotide polymorphisms (cSNPs) from 7,600 expressed sequence tags (ESTs) derived from three cDNA libraries (liver, M. longissimus dorsi, and intermuscular fat) of Hanwoo (Korean native cattle) steers. From the 7,600 ESTs, 829 contigs comprising more than two EST reads were assembled using the Phrap assembler. Based on the contig analysis, 201 candidate cSNPs were identified in 129 contigs, in which transitions (69%) outnumbered transversions (31%). To verify whether the predicted cSNPs are real, 17 SNPs involved in lipid and energy metabolism were selected from the ESTs. Twelve of these were confirmed to be real while five were identified as artifacts, possibly due to expressed sequence tag sequence error. Further analysis of the 12 verified cSNPs was performed using the program BLASTX. Five were identified as nonsynonymous cSNPs, five were synonymous cSNPs, and two SNPs were located in 3'-UTRs. Our data indicated that a relatively high SNP prediction rate (71%) from a large EST database could produce abundant cSNPs rapidly, which can be used as valuable genetic markers in cattle.

Application of machine learning in SNP discovery.

BackgroundSingle nucleotide polymorphisms (SNP) constitute more than 90% of the genetic variation, and hence can account for most trait differences among individuals in a given species. Polymorphism detection software PolyBayes and PolyPhred give high false positive SNP predictions even with stringent parameter values. We developed a machine learning (ML) method to augment PolyBayes to improve its prediction accuracy. ML methods have also been successfully applied to other bioinformatics problems in predicting genes, promoters, transcription factor binding sites and protein structures.ResultsThe ML program C4.5 was applied to a set of features in order to build a SNP classifier from training data based on human expert decisions (True/False). The training data were 27,275 candidate SNP generated by sequencing 1973 STS (sequence tag sites) (12 Mb) in both directions from 6 diverse homozygous soybean cultivars and PolyBayes analysis. Test data of 18,390 candidate SNP were generated similarly from 1359 additional STS (8 Mb). SNP from both sets were classified by experts. After training the ML classifier, it agreed with the experts on 97.3% of test data compared with 7.8% agreement between PolyBayes and experts. The PolyBayes positive predictive values (PPV) (i.e., fraction of candidate SNP being real) were 7.8% for all predictions and 16.7% for those with 100% posterior probability of being real. Using ML improved the PPV to 84.8%, a 5- to 10-fold increase. While both ML and PolyBayes produced a similar number of true positives, the ML program generated only 249 false positives as compared to 16,955 for PolyBayes. The complexity of the soybean genome may have contributed to high false SNP predictions by PolyBayes and hence results may differ for other genomes.ConclusionA machine learning (ML) method was developed as a supplementary feature to the polymorphism detection software for improving prediction accuracies. The results from this study indicate that a trained ML classifier can significantly reduce human intervention and in this case achieved a 5–10 fold enhanced productivity. The optimized feature set and ML framework can also be applied to all polymorphism discovery software. ML support software is written in Perl and can be easily integrated into an existing SNP discovery pipeline.