Straub, R. | bioRxiv Science

bioRxiv2024

scMeFormer: a transformer-based deep learning model for imputing DNA methylation states in single cells enhances the detection of epigenetic alterations in schizophrenia

DNA methylation (DNAm), a crucial epigenetic mark, plays a key role in gene regulation, mammalian development, and various human diseases. Single-cell technologies enable the profiling of DNAm states at cytosines within the DNA sequence of individual cells, but they often suffer from limited coverage of CpG sites. In this study, we introduce scMeFormer, a transformer-based deep learning model designed to impute DNAm states for each CpG site in single cells. Through comprehensive evaluations, we demonstrate the superior performance of scMeFormer compared to alternative models across four single-nucleus DNAm datasets generated by distinct technologies. Remarkably, scMeFormer exhibits high-fidelity imputation, even when dealing with significantly reduced coverage, as low as 10% of the original CpG sites. Furthermore, we applied scMeFormer to a single-nucleus DNAm dataset generated from the prefrontal cortex of four schizophrenia patients and four neurotypical controls. This enabled the identification of thousands of differentially methylated regions associated with schizophrenia that would have remained undetectable without imputation and added granularity to our understanding of epigenetic alterations in schizophrenia within specific cell types. Our study highlights the power of deep learning in imputing DNAm states in single cells, and we expect scMeFormer to be a valuable tool for single-cell DNAm studies.

genomics↗

bioRxiv2021

Specific BRCA and immune configurations determine optimal response to platinum-based chemotherapy in triple negative breast and ovarian carcinomas

Loss of homologous recombination repair (HRR) via germline and somatic BRCA1 or BRCA2 gene mutations and via BRCA1 promoter methylation has been associated with better response to platinum agents and PARP inhibitors, in both triple negative breast cancer (TNBC) and ovarian carcinoma (OvCa). A major conundrum arising from recent clinical studies is why cancers with BRCA1 promoter methylation (BRCA1meth) respond more poorly as compared to those bearing mutations in BRCA1 and BRCA2 (BRCAmut), given the biologically equivalent HRR deficiency in both states. We dissected this problem through detailed genomic analyses of primary TNBC and OvCa cohorts, as well as experimentation with patient-derived xenograft (PDX) models and genetically engineered cell lines. Using the precise genomic scar of the tandem duplicator phenotype as a precise genomic indicator of BRCA1 deficiency, we found that, in all cohorts, BRCA1mut and BRCA1meth cancers share an equivalent degree of BRCA1-linked genomic rearrangements. Nonetheless, we consistently found that patients with BRCAmut cancers, but not those with BRCA1meth cancers, had significantly better response outcomes when compared to those with BRCA proficient cancers. When fully promoter methylated BRCA1 PDX TNBCs were exposed to a single short course of platinum chemotherapy an unmethylated BRCA1 promoter allele emerged in resultant tumors associated with an increase in BRCA1 expression. A separate analysis of PDXs derived from treatment naive TNBCs featured complete methylation of the BRCA1 promoter, whereas those derived from post-chemotherapy TNBCs invariably had only partial methylation. PDXs with partial methylation were significantly associated with lower response rates to in vivo platinum-based therapy compared to those with complete promoter methylation. Using single cell clonal expansions from a partially BRCA1meth PDX, we confirmed that the reduced level of methylation was due to the demethylation of one of the BRCA1 promoter alleles and not to the outgrowth of a non-methylated clone. Clinically, analysis of primary OvCas confirmed that high levels of BRCA1 methylation were significantly associated with reduced BRCA1 gene expression whereas cancers with lower levels of BRCA1 methylation had expression levels approaching those found in BRCA1 proficient cancers. These data suggest that unlike BRCAmut cancers, where HRR deficiency is achieved via mutations that are genetically fixed, BRCA1meth cancers are highly adaptive to genotoxin exposure and more likely to recover BRCA1 expression, which may explain their poorer therapeutic response. We further found that an increased immune transcriptional signal, especially an elevated M1 macrophage signature, is associated with enhanced response to platinum-based chemotherapy only in patients with BRCA proficient cancers, in both TNBC and OvCa cohorts underscoring the importance of characterizing molecular heterogeneity to enhance predictive precision in assigning response probabilities in TNBC and OvCa.

cancer biology↗