Martin Kircher

Dr Martin Kircher
Junior Group Leader 'Computational Genome Biology’
Berlin Institute of Health

Dr. Martin Kircher joined BIH in March 2017 as a new Junior Research Group (JRG) Leader in the field of Bioinformatics. He describes the work of his "Computational Genome Biology" group and experience in a brief portrait.

The research of his group focuses on computational approaches of identifying functionally relevant genetic changes in disease and adaptation as well as developing more sensitive methods in diagnostics (especially exome, genome and cell-free DNA sequencing). Generally, our research spans the fields of sequence analysis, data mining, machine learning and functional genomics.

Based on a broad interest in genetics, epigenetics, and human adaptation, the group develops computational solutions to overcome technical and experimental obstacles in high-throughput sequencing-based protocols. The main focus area are computational approaches for identifying functionally relevant genetic.

Read more

Abstract

Interpretation of human genetic variation across the genome

The use of sequencing approaches for the identification of disease causal mutations is rapidly gaining traction in research and clinical settings, but the interpretation of the identified variants remains a major challenge. When scaling from exome to genome sequencing, the vast majority of variants fall in non-coding regions of the genome. However, we currently have a very limited toolset for their interpretation and almost no validation data for predictive approaches or training data that can be applied for machine learning strategies. In 2014, we developed a broadly applicable metric that objectively weights and integrates a large, diverse, and otherwise unwieldy collection of annotation data available. Combined Annotation Dependent Depletion (CADD, http://cadd.gs.washington.edu) integrates sequence annotations by contrasting variants that survived natural selection with simulated mutations. I will describe our general method as well as the integration of additional annotations and methodological improvements we made over the last years. In a second part, I will focus on how we create and use experimental data from massively parallel reporter assays (MPRAs) to assess the performance of currently available computational predictors of non-coding sequence. We used saturation mutagenesis MPRA to create comprehensive base-pair level activity maps of different promoter sequences previously implicated in human disease. With read outs of more than 10,000 single nucleotide alterations, we created an unprecedented database for the interpretation of potentially disease-causing regulatory mutations. I will show benchmarking results of commonly used computational tools and will highlight how performance is highly dependent on the genomic sequence under consideration.