Position kernels as a key to making sense of very rare and private single-nucleotide variants

Ulrich Bodenhofer, Sepp Hochreiter

Research output: Contribution to conferencePaperpeer-review

14 Downloads (Pure)


We present an approach for convolving single-nucleotide variants (SNVs) with a position kernel in order to augment SNVs with information about close-by SNVs. By means of the Position-Dependent Kernel Association Test (PODKAT), we demonstrate the potential of this approach to leverage the analysis of rare and private SNVs. Finally, we also provide some ideas how machine-learning based predictions from genomic data can benefit from this augmentation.
Original languageEnglish (American)
Publication statusPublished - Dec 2017
Externally publishedYes
EventNIPS Workshop on Machine Learning in Computational Biology - Long Beach, CA, United States
Duration: 9 Dec 20179 Dec 2017


WorkshopNIPS Workshop on Machine Learning in Computational Biology
Country/TerritoryUnited States

Cite this