This paper here was published in Bioinformatics in a cooperation with Prof. Harald Binder from the Institute of Medical Biometry and Statistics (IMBI) at the university of Freiburg, Germany.

His team developed a new approach to leverage deep learning with Boltzmann machines to identify clinically relevant patterns of single nucleotide polymorphisms (SNPs) in spite of the high dimensionality and low sample size typically associated with this type of data. We validated the set of candidate SNPs that they discovered using their new approach in an independent survival analysis of a cohort of leukemia patients from The Cancer Genome Atlas (TCGA) - they were indeed jointly associated with outcome.