The IMPACT of INDEL realignment: Detecting insertions and deletions longer than 30 base pairs with ABRA
2 likes790 views
The document discusses the limitations of existing indel realignment methods in detecting insertions and deletions longer than 30 base pairs in cancer genome analysis. It presents a comparison of two software tools, GATK's indel realigner and a new tool called Abra, highlighting how Abra improves indel detection and increases variant frequency accuracy. The findings emphasize the significance of effective indel detection in personalizing cancer treatment.
The IMPACT of INDEL realignment: Detecting insertions and deletions longer than 30 base pairs with ABRA
- 1. The IMPACT of INDEL realignment: Detecting insertions and deletions longer than 30 base pairs with ABRA Kirk Thaker1, Ronak Shah2, Michael Berger2 1Riverdale Country School, Bronx, NY, 2Department of Pathology, Memorial Sloan Kettering Cancer Center, New York, NY ABSTRACT Background Cancer is a disease of the genome –most of its forms result from a buildup of genetic alterations that, directly or indirectly, allow the patient’s cells to proliferate without restraint. For decades, identifying and targeting cancer mutations for treatment was impractical due to the limitations of sequencing technology. However, the rise of high-throughput next- generation sequencing (NGS) tools has allowed researchers to rapidly and cheaply sequence large, targeted regions of DNA. MSK-IMPACT(Memorial Sloan Kettering-IntegratedMutation Profiling of Actionable Cancer Targets), a sequencing platform with an associated computational pipeline, takes advantage of improvements in sequencing technology to analyzetumor specimensfor clinically actionable variants in341 cancer- associatedgenes.Criticalto IMPACT’s efficacy is the detection of somatic DNAalterationslike INDELs, which are insertions or deletions of nucleotides. Current sequence aligners have difficulty accuratelymapping reads (short, overlapping DNA sequences) containing morethan a single base change, let alone reads containing INDELs. This flaw necessitates the use of INDEL realigners, whichrearrange reads inregions where INDELs might exist in order to identify them more easily. Currently, the INDEL realignment software associated withMSK- IMPACT’scomputational pipeline, the Genome Analysis Toolkit’s IndelRealigner (GATK), canonly efficiently resolveINDELsshorter than 30 base pairs, which limits theplatform’sreliability forINDELdetection. Thus, wetested and compared the performance of a new INDEL realigner called ABRA (Assembly BasedRe-Aligner) to that of GATK’s IndelRealigner. Objectives 1.To resolve poorly aligned genomic regions caused by occurrence ofINDELsand repeat sequences. 2.To improve INDEL detection performancewith emphasis on both finding INDELslonger than 30 bp and on improving the accuracy of each INDELs variant frequency. CONCLUSIONS METHODS RESULTS •ABRA increased our confidence in already existing variant calls by increasing the variant frequency of the alternate allele. •ABRA detected INDELs longer than 30 base pairs, especially in regions that previously exhibited “messy” or unclear read alignments. •INDELs, unless they occur in multiples of 3, often negatively impact the structure of proteins they code -therefore, not identifying INDELs presents an obstacle in the creation of personalized cancer medicine. Processing BAMs: #of samples SNVs gained SNVs dropped INDELs gained INDELs dropped Total gained Total dropped 151 1 1 12 2 13 3 ABRA increases supporting evidence for already-existing INDEL calls ACKNOWLEDGEMENTS ABRA detects INDELs longer than 30 base pairs where GATK cannot I would like to thank Ronak Shah and Dr. Michael Berger for all of their instruction, support, and advice in making this project. Targeted sequencing Computational pipeline Prepare 24-48 libraries Probes for 340 cancer genes Sequence to 500- 1000X (HiSeq 2500) Hybridize & select (NimbleGen SeqCap: MSK-IMPACT Assay) Align to genome & analyze Raw data Mapping Processing BAMs Metrics calculations Variant calling, filtering, and genotyping Variant annotation and filtering Manual analysis Mark duplicates (PICARD tools) INDEL realignment Base quality recalibration GATK’s IndelRealigner ABRA Figure 1: GATK’s IndelRealigner attempts to minimize the number of mismatches, preferring deletions and insertions over individual SNPs Figure 2: ABRA creates a de Bruijn graph of k-mers(sequences) of variable lengths and maps back locally assembled reads using BWA-MEM. Table 1: After realignment with ABRA and GATK on a common set of mutations, we found that ABRA increased the variant frequency of 13 of those events, letting them pass our filters and be called as significant by the pipeline. Although GATK had already detected those events, ABRA increased our confidence in those calls, to the point where we could consider them meaningful. Figure 4: For the 10 pools (151 samples) above, ABRA (brown) consistently increased the variant frequency of INDELs also found by GATK’s IndelRealigner (green). Tumor VF=21% Normal VF=0% Tumor VF=4.3% Normal VF=0% ABRA GATK Figure 6: Above, a 21 bp insertion detected by both ABRA and GATK in the MAP3K1 gene. Here we clearly see that ABRA’s alignment exhibits a tumor variant frequency that is almost 5 times higher than that of GATK’s IndelRealigner. 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 16D 61 62 63 64 65 66 67 69 70 Variant frequency (decimal) Pool name Variant frequency of common INDELs between GATK and ABRA GATK VF ABRA VF Figure 8: A deletion and another mutation event called by GATK in the BARD1 gene was resolved far more cleanly by ABRA into a single deletion event with a separate SNV farther away. ABRA GATK Figure 9: Here, ABRA is able to detect a significant exon 11 insertion (45bp) and deletion (42bp) in the KIT gene, which is usually relevant for patients with gastrointestinal cancer. Tumor VF=41% Normal VF=40% Tumor VF=10% Normal VF=0% Figure 7: A 41 bp deletion called as a true positive with GATK because of no presence in the normal is found to be a sequencing artifact after applying ABRA as a realigner. ABRA GATK ABRA GATK ABRA resolves poorly aligned regions Tumor VF=25% Normal VF=0% Tumor VF=0% Normal VF=0% ABRA GATK Tumor VF=22% Normal VF=0% Tumor VF=0% Normal VF=0% Tumor VF=34% Normal VF=0% Tumor VF=0% Normal VF=0% ABRA presents a more parsimonious alignment of the sequencing data. ABRA GATK Figure 10: ABRA is able to resolve a large deletion (>100bp) in the HRAS gene, which can significantly impact patient treatment and prognosis in bladder cancer. Tumor VF=74% Normal VF=0% Tumor VF=0% Normal VF=0% local multiple sequence alignment The importance of IMPACT is twofold: it allows oncologists to better understand their patient’s disease and decide upon treatment and researchers can use it to retroactively analyze tumor specimens for common mutations