Misidentification of MLL3 and other mutations in cancer due to highly homologous genomic regions.
Timothy G BowlerKith PradhanYu KongMatthias BartensteinKerry A MorroneAshwin SridharanRachel M KesselAditi ShastriOrsi GiriczTushar D BhagatShanisha Gordon-MitchellMersedeh RohanizadeganLauren HoodaIshan DattBartlomiej P PrzychodzenSimrit ParmarShahina MaqboolJaroslaw P MaciejewskiUlrich SteidlJohn M GreallyAmit K VermaPublished in: Leukemia & lymphoma (2019)
The MLL3 gene has been shown to be recurrently mutated in many malignancies including in families with acute myeloid leukemia. We demonstrate that many MLL3 variant calls made by exome sequencing are false positives due to misalignment to homologous regions, including a region on chr21, and can only be validated by long-range PCR. Numerous other recurrently mutated genes reported in COSMIC and TCGA databases have pseudogenes and cannot also be validated by conventional short read-based sequencing approaches. Genome-wide identification of pseudogene regions demonstrates that frequency of these homologous regions is increased with sequencing read lengths below 200 bps. To enable identification of poor quality sequencing variants in prospective studies, we generated novel genome-wide maps of regions with poor mappability that can be used in variant calling algorithms. Taken together, our findings reveal that pseudogene regions are a source of false-positive mutations in cancers.
Keyphrases
- acute myeloid leukemia
- genome wide
- genome wide identification
- copy number
- single cell
- dna damage
- dna repair
- dna methylation
- machine learning
- transcription factor
- single molecule
- allogeneic hematopoietic stem cell transplantation
- squamous cell carcinoma
- protein protein
- papillary thyroid
- bioinformatics analysis
- quality improvement