Document Actions

Alexey Nesvizhskii(03-03-05)

Institute for Systems Biology

Title: Computational Challenges in Proteomics: From Protein Identification to Biological Interpretation

Abstract: Tandem mass spectrometry-based proteomic approach (shotgun proteomics) has been used increasingly for high throughput analysis of complex protein samples. Similar to other global high throughput technologies such as microarray gene expression analysis, proteomics is extremely dependent on the ability to quickly and reliably analyze large amounts of experimental data and convert it into biological knowledge. In this talk, I will describe a set of statistical models and computational tools that we developed in order to facilitate the global quantitative proteomics analysis. In particular, I will focus on the following three topics: statistical validation of large-scale peptide and protein identification datasets, modeling of quantitative protein expression profiles to determine the composition of macromolecular complexes or cell organelles, and application of mass spectrometry-derived data to the whole genome with the purpose of gene validation or elucidation of global patterns of protein expression.

Comment:

Bayesian classification. Lots of prior and posterior probabilities. Calculate prior probabilites from current database data. EM mixture model training to get the parameters. Combine multiple probabilities, $p=1-\sum(1-p_i)$ . BIC(Bayesian information criteria), AIC information criteria. Occam's razor approach.

He presented an alternative to p-value and expectation value, scoring.

Yu Huang 2006-03-25

November 2009

Yu S. Huang Homepage, USC MCB/CMB

Sections

Personal tools

Document Actions

Alexey Nesvizhskii(03-03-05)