Recent News
New AI tool for detecting de novo mutational signatures
We developed a novel machine learning tool, SigProfilerExtractor, that allows extracting de novo mutational signatures from cancer genomic data. We demonstrated that SigProfilerExtractor outperforms 13 existing approaches, by conducting the largest benchmarking of bioinformatics tools for extracting mutational signatures using more than 80,000 synthetic cancer samples. Applying SigProfilerExtractor to more than 23,000 sequenced human cancers revealed novel insights about the processes leading to somatic mutation accumulation and tumorigenesis, including a novel mutational signature related to tobacco smoking mutagenesis in bladder cancer and normal bladder epithelium.
Copy number signatures in human cancer
As part of a large collaborative project, our lab co-led the development of a conceptual framework for examining the patterns of copy number alterations in human cancer. This conceptual framework allowed deriving the first reference set of copy number signatures across the spectrum of human neoplasia. The performed analysis revealed 21 distinct copy number signatures across 9,873 cancers representing 33 human cancer types. Moreover, for some copy number signatures, we were able to elucidate their biological mechanisms and to show their immediate clinical relevance. The reference set of copy number signatures was deposited on the COSMIC database and can be freely and easily utilized by other researchers as part of their omics studies.
Clustered mutations as important players in cancer evolution
Our analysis revealed that clustered mutations play a key role in driving cancer evolution and can be utilize as clinical biomarkers for predicting patient survival. Importantly, a class of clustered mutations, which we termed kyklonas (Greek for cyclone), were shown to be generated by the activity of APOBEC3 deaminases on extrachromosomal DNA, small circular pieces of DNA with a structure and replication pattern similar to that of a virus. APOBEC3 likely mistakes ecDNA as a virus and tries to eliminate it. But repeated mutagenic attacks to ecDNA result in clustered mutagenesis of the oncogenes it contains. Rather than eradicating the threat, this mutagenesis contributes to tumor evolution, promoting aggressive tumor behavior and treatment resistance.
Unravelling the mystery of esophageal cancer incidence
As part of the global Mutographs project, we recruited patients from eight countries around the world with drastically different risk for developing esophageal squamous cell carcinoma. When we embarked on this study, we were looking for a genomic mutational signature which was only present or increased in areas where people have high risk for esophageal squamous cell carcinoma. The analysis of esophageal cancer genomes, using our advanced machine learning approaches, revealed an unexpected surprise - a common mutational process, driven by the APOBEC3 family of deaminases, is the ubiquitous culprit causing cancer in almost all patients in these eight countries.
Understanding the cause of lung cancer in never smokers
As part of the global Scherlock-lung project, we investigated the molecular mechanisms by which people, who never smoked, get lung cancer. Our analysis revealed that the lung cancer genomes of never smokers are dramatically different when compared to the lung cancer genomes of tobacco smokers. Moreover, our examination revealed strong evidence for difference molecular subtypes within lung cancers of never smokers. These subtypes can be leveraged as a clinical utility for better managing and treating lung cancer in patients, who have never smoked.