Clàudia Serrano Colomé

The paper identifies a hypermutation hotspot at transcription start sites in the human germline, linked to early embryogenesis and distinct from known mutagenic processes.

The paper shows non-hypermutator cancers adapt by reversing germline mutation bias, enabling access to driver mutations through altered mutation spectra.

This paper extends the Walsh-Hadamard transform to model epistasis in genetic landscapes with multiple alleles per position, enabling analysis of pairwise and higher-order interactions in simulated and empirical data.

The paper introduces SigNet, a deep learning method for mutational signature decomposition, showing improved accuracy and robustness in estimating signature contributions in cancer genomes.