Utilizing insights of DNA repair machinery to discover MMEJ deletions and novel mechanisms

We developed Del-read, an algorithm targeting medium-sized deletions (6-100 bp) in short-reads, which are challenging for current variant callers relying on alignment. Our focus was on Micro-Homolog mediated End Joining deletions (MMEJ-dels), prevalent in myeloid malignancies. MMEJ-dels follow a distinct pattern, occurring between two homologies, allowing us to generate a comprehensive list of MMEJ-dels in the exome. Using Del-read, we identified numerous novel germline and somatic MMEJ-dels in BEAT-AML and TCGA-breast datasets. Validation in 672 healthy individuals confirmed their presence. These novel MMEJ-dels were linked to genomic features associated with replication stress, like G-quadruplexes and minisatellite. Additionally, we observed a new category of MMEJ-dels with an imperfect-match at the flanking sequences of the homologies, suggesting a mechanism involving mispairing in homology alignment. We demonstrated robustness of the repair system despite CRISPR/Cas9-induced mismatches in the homologies. Further analysis of the canonical ASXL1 deletion revealed a diverse array of these imperfect-matches. This suggests a potentially more flexible and error-prone MMEJ repair system than previously understood. Our findings highlight Del-read's potential in uncovering previously undetected deletions and deepen our understanding of repair mechanisms.