BEGIN:VCALENDAR VERSION:2.0 PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4// BEGIN:VEVENT UID:20250505T134614EDT-2515CiUcOS@132.216.98.100 DTSTAMP:20250505T174614Z DESCRIPTION:Getting more out of mass spectrometry-based proteomics using su pervised learning approaches and on-the-fly data analysis\n\nMathieu Laval lee (University of Ottawa)\n Tuesday February 5\, 12-1pm\n Montreal Neurolog ical Institute\, deGrandpre Communications Centre\n \n Abstract: Mass spectr ometry-based proteomics is widely used to identify proteins in complex bio logical samples. Current proteomics approaches generate hundreds of thousa nds of mass spectra\, yet\, on average\, only 25% of the mass spectra acqu ired in a mass spectrometry experiment are computationally matched to prot ein sequences. Furthermore\, since this computational matching typically t akes place after mass spectrometry data acquisition\, many abundant protei ns are analyzed in excess than what is necessary for a confident identific ation\, leaving little mass spectrometry time for the analysis of lower ab undance proteins. Increasing protein identification sensitivity is critica l to provide a comprehensive understanding of the underlying biology of co mplex samples. Protein-protein interactions contain information that can i mprove protein identification rate in mass spectrometry\; information that is not used by most current algorithms. We therefore propose a novel mach ine learning algorithm that assesses the confidence of protein identificat ions using mass spectrometry data features and confidence scores along wit h protein-protein interaction data. Our approach is based on the hypothesi s that the confidence of the identification of a given protein P in a samp le increases when proteins interacting with P are also observed in the sam e sample. Upon benchmarking against a state-of-the-art approach\, our algo rithm identifies more spectra\, peptides and proteins at low false discove ry rates. Also\, to improve identification sensitivity of low abundance pr oteins\, we designed a machine learning classifier that evaluates the reli ability of protein identifications on the fly\, as mass spectra are acquir ed. Proteins that are deemed confidently identified are excluded from furt her analysis in real-time\, saving mass spectrometry resources for lower a bundance proteins. We show in silico that our approach can identify a simi lar number of proteins using significantly less mass spectrometry time tha n a traditional proteomics analysis\, thereby freeing resources for more p rotein identifications. Finally\, our algorithms improve our ability to id entify proteins in complex samples and will provide a more comprehensive u nderstanding of the biological mechanisms of the cell.\n DTSTART:20190205T170000Z DTEND:20190205T180000Z LOCATION:deGrandpre Communications Centre\, Montreal Neurological Institute \, CA\, QC\, Montreal\, H3A 2B4\, 3801 rue University SUMMARY:Seminar Series in Quantitative Life Sciences and Medicine URL:/qls/channels/event/seminar-series-quantitative-li fe-sciences-and-medicine-293647 END:VEVENT END:VCALENDAR