Notes on #ISMBECCB Sunday Afternoon Sessions

Mapping the Strategies of Viruses Hijacking Human Host Cells: An Experimental and Computational Comparative Study by Jacques Colinge

Original Paper: Viral immune modulators perturb the human molecular network by common and unique strategies. Pichlmair et al. (2012)

Interesting use of tagged proteins with purification and mass spec to analyse the interactions between viral and host proteins. Many viruses target the same proteins, but there is specificity too.

Drilling down on HCV, USP19 is a major target of viral proteins. This protein promotes degradation of misfolded proteins, but it may also be able to detect non-host proteins in general.

Human proteins that are targeted by virusses are much more likely to be hubs of protein/protein interaction networks. As virus are often very small so can only target a small number of proteins, thus targetting the hubs may be efficient (at the cost of specificity of effect).

Multi-task learning for Host-Pathogen protein interactions by Meghana Kshirsagar

Many pathogens will use the same strategies, thus we can learn across different pathogens to find similarities between different host/pathogen pairs. Their working hypothesis is that pathogens will target the same pathways in the host.

This is formalised as a multi-task learning process: loss on the training set is regularized by the difference of pairwise pathway signatures. Technically, this is nice because it can incorporate unlabeled data or missing data (as long as you can compute signatures).


Notes on #ISMBECCB Highlights Session (Sunday Morning)

(I missed the first half of the first talk, so I won’t include it. Also, the internet is not good enough for me to get all the links. Sorry)

Of Men and Not Mice: Comparative Genomic Analysis of Human Diseases and Mouse Models by Wenzhong Xiao

Wenzhong Xiao presented an empirical study of correlation between immune response of mice and men. The correlations were very low, which is a warning to be careful in interpreting animal models results. Money quote: “Mice are not human. There are several reasons for that.”

An audience member raised the possibility of using humanized models, which was a great point. I’ll add that the immune system and immune system dysfunction may be where mice and men differ the most and results there do not invalidate results in other areas of study.

Impact of genetic dynamics and single-cell heterogeneity on development of nonstandard personalized medicine strategies for cancer by Chen-Hsiang Yeang

Simulation study of using different strategies for cancer treatment in the present of resistant mutations. “The current system is often like a greedy algorithm: do X until resistance to X emerges, switch to Y. Repeat. Better strategies are possible.”

Very interesting points come out of simple models, but it felt like the start of a conversation rather than an answer.

Interesting presentation detail: author used references to video games as one would use references to literature 100 years ago.

Systems-based metatranscriptomic analysis by Xuejian Xiong.

Original Paper: He, D., Miao, M., Sitarz, E.E., Muiznieks, L.D., Reichheld, S., Stahl, R.J., Keeley, F.W. and Parkinson, J. (2012) Polymorphisms in the Human Tropoelastin Gene Modify in vitro Self-Assembly and Mechanical Properties of Elastin-like Polypeptides. PLoS ONE. 7(9): e46130

Study on non-obese diabetic mice with Illumina sequencing. They projected their reads into enzyme space to perform analysis at the metabolic network level.

Interesting technical points: they use alignment in peptide space instead of nucleotide space to get around variability in codon encoding. They also found that Trinity worked best for their data.