No abstract

The human genome contains over 100 million SNPs, most of which are C/T (G/A) variations. The type and sequence context of these SNPs are not random, suggesting that they are caused by distinct mutational processes. Deciphering the mutational signatures is a crucial step to discovering the molecular processes responsible for DNA variations across human populations, and potentially for causing genetic diseases. Our analyses of the 1000 Genomes Project SNPs and germline de novo mutations suggest that at least four mutational processes are responsible for human genetic variations. One process is European-specific and no longer active. The remaining three processes are currently active in all human populations. Two of the active processes co-occur and leave a single joint mutational signature in human nuclear DNA. The third active process is specific to mitochondrial DNA, and inflicts C-to-T mutations at mostly non-CG sites. We found neither evidence of APOBEC-induced cytosine deamination in the human germline, nor de novo mutation enrichment within certain regions of the human genome.

DNA of many breast tumors is barraged by C-to-T/G mutations within TCW (W:T,A). These mutations are attributed to the aberrant expression and activity of APOBEC3 enzymes. They have been shown to account for many driver mutations in genes such as PIK3CA, ERBB2, and PPP2R1A, however their precise source and also their roles in tumor development, evolution, and patient survival are debated. Currently, quantification of APOBEC3 expression changes in tumor cells is confounded by the ubiquitous expression of these enzymes in immune infiltrating cells. In this study, we used a novel quantitative biology approach to determine the expression profiles of APOBEC3 enzymes in breast tumor and tumor microenvironment cells from >1,000 patients. We combined diverse datasets including tumor/matched normal RNAseqs, tumor somatic mutations, cell line RNAseqs and mutations, estimates of tumor purities and immune cell compositions, and expression of purified cell populations to show that in breast cancer there is only a single APOBEC3 dysregulation process. This process is subtype-independent and is represented by APOBEC3B upregulation and and extreme APOBEC3C downregulation. Compared to all other tumor types, breast tumors are affected the most by this process. Citation Format: Hamid Hamidi, Azad Khosh, Hamzeh Rahimi, Diako Ebrahimi. Profiling of APOBEC3 dysregulation in breast cancer [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P5-13-05.

Deep learning has been widely considered more effective than traditional statistical models in modeling biological complex data such as single-cell omics. Here we show the devil is hidden in details: by adapting a modern gradient solver to a traditional linear mixed model, we showed that conventional models can outperform deep models in terms of both speed and accuracy. This work reveals the potential of re-implementing traditional models with modern solvers.