Dr. Käll has a PhD from Karolinska Institutet and completed a post-doc at the University of Washington. His group develops methods for facilitating the interpretation of high-throughput experiments. He has developed a wide variety of bioinformatical tools based on machine learning techniques, such as Phobius/Philius, MaRaCluster, and Percolator.

Shotgun proteomics is a powerful technique to identify and quantify many proteins from complex mixtures through mass spectrometry. However, in label-free shotgun proteomics, quantifying proteins accurately presents many challenges, stemming from numerous error sources. Traditional methods to pinpoint differential proteins often employ filters at intermediate steps to manage error rates. Yet, these approaches might inadvertently neglect certain errors and treat filtered lists as unequivocally accurate in subsequent stages, leading to potential miscalculations in error rates.

Dr.Käll's talk will delve into an innovative solution called Triqler - a probabilistic graphical model designed to tackle these issues head-on. Unlike conventional methods, Triqler emphasizes error propagation throughout all stages, valuing the use of distributions over mere point estimates. Instead of bypassing uncertainties, Triqler reveals and underscores them by providing posterior probabilities for fold changes between different treatment groups. By advancing the precision and accuracy of data analysis, the Triqler method substantially enhances our ability to glean meaningful biological insights from proteomics studies, driving our understanding of complex biological systems.

Talk Title: Statistical Analysis of Label-free Shotgun Proteomics Data

Hosts: Andrea Fossati