Optimizing microsatellite instability detection in cancer with MANTIS
Loading...
Date
2017-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact Journals, LLC
Abstract
Purpose
Microsatellites are genomic loci that contain several repeats of short (1-6 base pair) nucleotide sequences. Microsatellite instability (MSI) is a genetic phenomenon in which uncorrected “slippage” of DNA fragments during DNA replication causes the repeat count to vary with cell division. MSI occurs frequently in several human cancers, most commonly colorectal, endometrial, and gastric adenocarcinoma. MSI results from defective DNA mismatch repair (MMR), which can occur either sporadically, or in patients with genetic predispositions. MSI is currently detected in these cancer types using two standard tests: immunohistochemistry (of four MMR proteins) and MSI-PCR (of five microsatellite loci). However, studies have provided increasing evidence of a potential role of MSI in several other cancer types, which may be missed with current standard diagnostic methods. The implementation of next generation sequencing (NGS) may facilitate ways to characterize the MSI status of multiple cancer types, and potentially identify MSI missed by current methods. We therefore sought to:
1) develop a new NGS-based method that utilizes matched tumor-normal data,
2) compare existing computational MSI detection methods, and
3) determine whether the number of loci evaluated has an impact on performance.
Research Method
We developed MANTIS, a novel computational method for research and clinical MSI detection in NGS data from matched tumor and normal samples. MANTIS was compared to the existing NGS-based algorithms mSINGS and MSISensor. For testing and validation, we applied all three tools to whole-exome sequencing data from 275 matched tumor and normal samples with known MSI status (from three cancer types) from The Cancer Genome Atlas. We then evaluated the tools’ performance in three additional cancer types, as well as their performance when utilizing variable numbers of target loci ranging from 10 to 2539.
Findings
All three computational methods were found to be accurate for detecting MSI, with MANTIS demonstrating the highest overall sensitivity (97.18%), specificity (99.68%) and accuracy (98.91%). MANTIS demonstrated superior performance to mSINGS and MSISensor in colorectal adenocarcinoma, gastric adenocarcinoma and pancreatic adenocarcinoma, and was noninferior in esophageal carcinoma and uterine carcinosarcoma. MSISensor performed slightly better in endometrial carcinoma.
We additionally calculated the most predictive microsatellite loci for each tool with each cancer type. Each tool was then evaluated using its best 10 loci, 20, 30, 40, 50, 100, 250, 500, 1000, and the full set of 2539 microsatellite loci. We found that, in general, MSISensor performance increases with more loci considered, and mSINGS performance increases with fewer loci. However, MANTIS performance remains consistent across the range of loci numbers tested, even with only 10 loci, and across cancer types. We also identified loci that were preferentially more unstable across different cancer types.
Implications
This study may facilitate clinical NGS-based testing for multiple cancer types that are not routinely tested for MSI, and identification of MSI-positive tumors can provide eligibility for novel immunotherapy trials and treatments. Our results indicate that MANTIS may be useful in conjunction with a small targeted NGS panel, for identifying MSI in tumors that are sequenced for other purposes, and in a variety of cancer types.
Description
Biological Sciences: 3rd Place (The Ohio State University Edward F. Hayes Graduate Research Forum)
Keywords
Microsatellite instability, Computational biology, Next-generation sequencing
Citation
Published citation: Kautto EA, Bonneville R, Miya J, et al: Performance evaluation for rapid detection of pan-cancer microsatellite instability with MANTIS. Oncotarget 8:7452-7463, 2016. DOI: 10.18632/oncotarget.13918