Human Proteome Variation — Rasmussen Group
The Rasmussen group focus on proteome and genome variation, coding variation and deep learning for integration of genomics, proteomics and clinical data.
The Rasmussen Group focus on computational analysis of human proteome variation. Currently proteome variation is primarily studied using genome and exome sequencing data, from which impact on protein structures and functionality can be investigated. The group aims at building infrastructure around human proteome variation as genome graph representations that will be used to store and analyse thousands to millions of human genomes.
Here an important challenge is to translate individual genome variation to functional proteome variation and the group works towards this by studying differential prediction of protein features such as secondary structure, cleavage events and post translational modifications. Additionally, this allows population-based fractional breakdowns on versions of the same protein as well as correlations between the main versions of different proteins in the same individual.
Second, the group focus on using deep learning for data integration and interpretation. The group aims at applying large scale deep learning to increase understanding of human diseases and use this to pave the way for more precise diagnostics, prevention and new treatment strategies. Here, we develop computational approaches for investigating the human microbiome and human diseases, and in particular to integrate massive genomics, metabolomics, proteomics, registry, wearable, speech and clinical data.
Finally, the group has a long standing interest in studying the evolution of human pathogens. Here, we analyze ancient DNA to identify proteome variation through time. This enables us to identify important protein changes in the evolution of deadly human pathogens from non-pathogenic ancestors.
Using population wide genome sequencing we have been involved in analysis of data from thousands of human genomes including the Danish pan genome, large world wide datasets and ancient genomes. Additionally, the group has discovered a previously unknown pandemic of plague in the Stone and Bronze Age shaping the genome of the modern European population and detailing the proteome evolutionary history of plague. Finally, within deep learning we have developed the first application of unsupervised deep learning within metagenomics and analysis of the human gut microbiome. Here, our method has improved binning of microbial genomes by up to 800%.