Pan cancer patterns of allelic imbalance from chromosomal alterations in 33 tumor types

Abstract

Somatic copy number alterations (SCNAs) serve as hallmarks of tumorigenesis and often result in deviations from one-to-one allelic ratios at heterozygous loci, leading to allelic imbalance (AI). The Cancer Genome Atlas (TCGA) reports SCNAs identified using a circular binary segmentation algorithm, providing segment mean copy number estimates from single-nucleotide polymorphism DNA microarray total intensities (log R ratio), but not allele-specific intensities (“B allele” frequencies) that inform of AI. Our approach provides more sensitive identification of SCNAs by modeling the “B allele” frequencies jointly, thereby bolstering the catalog of chromosomal alterations in this widely utilized resource. Here we present AI summaries for all 33 tumor sites in TCGA, including those induced by SCNAs and copy-neutral loss-of-heterozygosity (cnLOH). We identified AI in 94% of the tumors, higher than in previous reports. Recurrent events included deletions of 17p, 9q, 3p, amplifications of 8q, 1q, 7p, as well as mixed event types on 8p and 13q. We also observed both site-specific and pan-cancer (spanning 17p) cnLOH, patterns which have not been comprehensively characterized. The identification of such cnLOH events elucidates tumor suppressors and multi-hit pathways to carcinogenesis. We also contrast the landscapes inferred from AI- and total intensity-derived SCNAs and propose an automated procedure to improve and adjust SCNAs in TCGA for cases where high levels of aneuploidy obscured baseline intensity identification. Our findings support the exploration of additional methods for robust automated inference procedures and to aid empirical discoveries across TCGA.

Chromosomal imbalances detected via RNA-sequencing in 28 cancers

“…To identify genomic regions that harbor SCNAs, we applied hapLOHseq. To assess the accuracy and potential of this approach, we compare SCNAs inferred from RNA-seq to high-confidence SCNA calls detected from DNA SNP microarray data, which have been documented previously ( Sivakumar et al , 2021 ) with an estimated false-positive rate <3% ( Vattathil and Scheet, 2016 ). For these purposes, since the RNA and DNA were derived from the same tissue (or tumor), we treat the SNP microarray results as a gold standard.…”

“… Note : hapLOHseq and CaSpER performance evaluation. Rows 1–3 show performance results at the gene level obtained by comparing each method to the gold standard ( Sivakumar et al , 2021 ). …”

“…Another approach for detection of SCNAs from bulk RNA profiling integrated coverage data and tumor-specific SCNA frequency patterns from public, external, data to identify chromosome-arm level aneuploidy, which was in turn assessed for association with prostate cancer outcomes ( Stopsack et al , 2019 ). Yet, these methods do not utilize haplotype information (the genetic makeup of a single chromosome that is passed on from a parent), which has been shown to increase power for SCNA detection in studies with SNP microarray data ( Baugher et al , 2013 ; Loh et al , 2018 ; Sivakumar et al , 2021 ; Vattathil and Scheet, 2013 ).…”

Oncogenic Potential of the Dual-Function Protein MEX3A

“…Despite epigenetic alterations like altered methylation pattern, as proposed for MEX3A in the context of breast cancer [ 58 ], genomic instability events come progressively in the focus of cancer biology, as chromosomal alterations are prominent in nearly all cancers and serve as the hallmark of cancer [ 59 , 60 ]. Analyses of somatic copy number alterations (SCNAs) revealed that roughly 94% harbor an allelic imbalance.…”

“…Analyses of somatic copy number alterations (SCNAs) revealed that roughly 94% harbor an allelic imbalance. In this context, amplification of chromosome 1q was amongst recurrent events in 33 tumor types [ 60 ], including hepatocellular carcinoma with a frequency of 46–86% [ 61 ]. MEX3A is encoded on chromosome 1q22 within a region (1q21-1q25) that is gained in various cancers [ 62 , 63 , 64 ] and encodes several pro-oncogenic genes, e.g., RAB25 [ 65 ].…”