Seminar by Mauno Vihinen

Speaker: Professor Mauno Vihinen, Protein Structure and Bioinformatics Group, Lund University
Host: Professor Kresten Lindorff-Larsen, Biomolecular Sciences, BIO-UCPH

Genetic variations can be efficiently be identified with powerful next generation sequencing methods. The bottleneck has moved from generating sequence data to interpreting this. The problem is especially difficult in medical genetics where the information can be used for diagnostic purposes, however, requirements for the quality and reliability of the interpretations are very high.

Computational approaches are widely used since already a single human exome contains some 10 000 variants causing amino acid substitutions, and many other types.

We have developed several artificial intelligence/machine learning tools to predict the impact of variations which can be broadly grouped as generic tools to predict the tolerance of variants, locus-specific tools to predict the tolerance of variants in specific proteins, genes or domains, and mechanism-specific tools to predict the mechanism of variation effect. The generic and locus-specific tools rank and prioritize variants and the mechanism-specific tools help to interpret the mechanism of variation effect.

PON-P2 is a generic tool for predicting variant pathogenicity with highest performance among related tools. The method computes reliability for each variation and classifies the reliably predicted variants as pathogenic or neutral and the unreliably predicted variants as unclassified.

Once harmful variants are identified we can address the mechanism or cause behind harmful ones. For this purpose we have developed several dedicated tools. Variants affecting protein structures can be investigated by analyzing effects on protein structure e.g. to investigate sterical clashes, electrostatic changes, modifications of binding sites etc.

The next layer is to investigate effects to cellular networks and pathways with systems biological tool set. The presented approaches will be computer-based, I plan to show real life examples on how to interpret consequences, effects and mechanisms of variants.