Uncovering missed indels by leveraging unmapped reads

Mohammad Shabbir Hasan,Xiaowei Wu,Liqing Zhang

doi:10.1038/s41598-019-47405-z

Abstract

In current practice, Next Generation Sequencing (NGS) applications start with mapping/aligning short reads to the reference genome, with the aim of identifying genetic variants. Although existing alignment tools have shown great accuracy in mapping short reads to the reference genome, a significant number of short reads still remain unmapped and are often excluded from downstream analyses thereby causing nonnegligible information loss in the subsequent variant calling procedure. This paper describes Genesis-indel, a computational pipeline that explores the unmapped reads to identify novel indels that are initially missed in the original procedure. Genesis-indel is applied to the unmapped reads of 30 breast cancer patients from TCGA. Results show that the unmapped reads are conserved between the two subtypes of breast cancer investigated in this study and might contribute to the divergence between the subtypes. Genesis-indel identifies 72,997 novel high-quality indels previously not found, among which 16,141 have not been annotated in the widely used mutation database. Statistical analysis of these indels shows significant enrichment of indels residing in oncogenes and tumour suppressor genes. Functional annotation further reveals that these indels are strongly correlated with pathways of cancer and can have high to moderate impact on protein functions. Additionally, some of the indels overlap with the genes that do not have any indel mutations called from the originally mapped reads but have been shown to contribute to the tumorigenesis in multiple carcinomas, further emphasizing the importance of rescuing indels hidden in the unmapped reads in cancer and disease studies.

Highlights

In current practice, Generation Sequencing (NGS) applications start with mapping/aligning short reads to the reference genome, with the aim of identifying genetic variants
These unmapped reads are not used for variant calling and downstream analyses, and mutations harboured in these unmapped reads remain hidden from any inference on important phenotype and/or their associations with any disease such as cancer
This paper emphasizes the interest of studying unmapped reads to cope with potential loss of important information and describes Genesis-indel, a computational pipeline to rescue novel high-quality indels by exploring unmapped reads that are normally discarded from the downstream analysis

Summary

Introduction

Generation Sequencing (NGS) applications start with mapping/aligning short reads to the reference genome, with the aim of identifying genetic variants. Many alignment algorithms have been developed to map the short reads to the reference genome, including MAQ4, SOAP5, BWA6, Bowtie[7], Bowtie[28], SNAP9, and SOAP210, to name a few These alignment tools are very efficient in aligning the short reads, a nonnegligible fraction of reads are left unmapped due to (1) structural variants longer than the allowed number of gaps and mismatches by the mapper, (2) sequencing error, or (3) sample contamination[11]. In current practice, these unmapped reads are not used for variant calling and downstream analyses, and mutations harboured in these unmapped reads remain hidden from any inference on important phenotype and/or their associations with any disease such as cancer. This study shows great promise in complementing the current procedure of read alignment and variant calling, shedding light on understanding the underlying mechanism of cancer progression and will be useful for clinical decision making

Methods

Results

Conclusion