Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2009 Sep;19(9):1542-52.
doi: 10.1101/gr.092072.109. Epub 2009 Jul 15.

A probabilistic approach for SNP discovery in high-throughput human resequencing data

Rose Hoberman  1 , Joana Dias, Bing Ge, Eef Harmsen, Michael Mayhew, Dominique J Verlaan, Tony Kwan, Ken Dewar, Mathieu Blanchette, Tomi Pastinen
Affiliations

A probabilistic approach for SNP discovery in high-throughput human resequencing data

Rose Hoberman et al. Genome Res. 2009 Sep.

Abstract

New high-throughput sequencing technologies are generating large amounts of sequence data, allowing the development of targeted large-scale resequencing studies. For these studies, accurate identification of polymorphic sites is crucial. Heterozygous sites are particularly difficult to identify, especially in regions of low coverage. We present a new strategy for identifying heterozygous sites in a single individual by using a machine learning approach that generates a heterozygosity score for each chromosomal position. Our approach also facilitates the identification of regions with unequal representation of two alleles and other poorly sequenced regions. The availability of confidence scores allows for a principled combination of sequencing results from multiple samples. We evaluate our method on a gold standard data genotype set from HapMap. We are able to classify sites in this data set as heterozygous or homozygous with 98.5% accuracy. In de novo data our probabilistic heterozygote detection ("ProbHD") is able to identify 93% of heterozygous sites at a <5% false call rate (FCR) as estimated based on independent genotyping results. In direct comparison of ProbHD with high-coverage 1000 Genomes sequencing available for a subset of our data, we observe >99.9% overall agreement for genotype calls and close to 90% agreement for heterozygote calls. Overall, our data indicate that high-throughput resequencing of human genomic regions requires careful attention to systematic biases in sample preparation as well as sequence contexts, and that their impact can be alleviated by machine learning-based sequence analyses allowing more accurate extraction of true DNA variants.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
Graphical overview of the four steps in the prediction pipeline. (1) Sequencing: Target regions are amplified by LR-PCR; amplicons are sequenced using a 454 GS-FLX sequencer. A set of sequence reads is generated by the 454 GS-FLX base-caller. (2) Alignment: Reads are aligned to the reference sequence and combined into a multiple sequence alignment (MSA). (3) Feature extraction: Numerical features are computed from the MSA for each site in the target region. (4) Training: Given a training set of sites with known genotypes from the HapMap database, we train a classifier to identify heterozygous sites from sequencing data. This classifier is then applied to novel data sets to identify novel SNPs.
Figure 2.
Figure 2.
(A) Average depth of coverage obtained for each successfully amplified LR-PCR fragment. Regions covered by two overlapping amplicons in the tiling path were excluded. (B) Frequency and magnitude of amplification bias, for amplicons with at least four known heterozygous sites, and average read coverage of at least 5×.
Figure 3.
Figure 3.
(A) Ability of three classifiers to identify known heterozygous sites. False call rate (FCR) (the fraction of called heterozygous SNPs that are known to be homozygous) is shown as a function of sensitivity (the fraction of known heterozygous sites called by each classifier). Three classifiers are compared: (1) GigaBayes; (2) ProbHD local-feature classifier, which considers all local features that could be extracted from the 454 GS-FLX generated MSA and quality scores file; and (3) ProbHD full classifier, which considers both local- and amplicon-level features from alignments generated by hAlign. (B) Estimated sensitivity and FCR for calling a site heterozygous, corrected for HapMap errors. (Dashed line) Assumes a HapMap error rate at the upper end of the 95% confidence interval; (dotted line) lower end; (solid line) no HapMap errors.
Figure 4.
Figure 4.
(A) Effect of coverage depth on prediction of known heterozygous sites. Sensitivity and positive predictive value (PPV, equal to 1 − FCR) are shown as a function of average depth of coverage. ProbHD results are shown with two different probability cutoffs for predicting heterozygous sites. A cutoff of c = 0.92 yields a conservative predictor that makes few false-positives, and a cutoff of c = 0.5 yields a very liberal predictor with higher sensitivity but higher FCR. Results are not corrected for HapMap errors. (B) Estimated de novo SNP-calling sensitivity and FCR, assuming 0.1% of sites are heterozygous. Well-sequenced sites are those sites with at least 13× coverage that are located on amplicons with minor allele deriving at least 25% of reads. The pronounced “elbow” is due to the severe imbalance between heterozygous and homozygous sites. Using a very conservative confidence threshold yields an error rate close to zero. However, as the threshold is lowered the percentage of homozygous sites miscalled as heterozygous sites eventually becomes nonzero. Even when the percentage of errors is quite small, the absolute number of errors quickly becomes large in comparison to the number of true hets, and the FCR climbs rapidly.
Figure 5.
Figure 5.
Common hets prediction. (A) Test set FCR (the percentage of predicted sites in HapMap that are not common hets) as a function of sensitivity. (B) The total number of predicted common hets as a function of desired sensitivity, based on one to four samples.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Albert TJ, Molla MN, Muzny DM, Nazareth L, Wheeler D, Song X, Richmond TA, Middle CM, Rodesch MJ, Packard CJ, et al. Direct selection of human genomic loci by microarray hybridization. Nat Methods. 2007;4:903–905. - PubMed
    1. Altshuler D, Daly MJ, Lander ES. Genetic mapping in human disease. Science. 2008;322:881–888. - PMC - PubMed
    1. Breiman L. Random forests. Mach Learn. 2001;45:5–32.
    1. Brockman W, Alvarez P, Young S, Garber M, Giannoukos G, Lee WL, Russ C, Lander ES, Nusbaum C, Jaffe DB. Quality scores and SNP detection in sequencing-by-synthesis systems. Genome Res. 2008;18:763–770. - PMC - PubMed
    1. Gnirke A, Melnikof A, Maguire J, Rogov P, LeProust EM, Brockman W, Fennel T, Giannoukos G, Fisher S, Russ C, et al. Solution hybrid selection with ultra-long oligonucleotides for massively parallel targeted sequencing. Nat Biotechnol. 2009;27:182–189. - PMC - PubMed

Publication types

LinkOut - more resources