*CANCELLED* - Distinguished Speaker Series: Charting γ secretase substrates by explainable AI
Prof. Dmitrij Frishman, Department of Bioinformatics, Technical University of Munich, Germany.
*CANCELLED*
Edmond J. Safra Center for Bioinformatics
Jointly with
School of Neurobiology, Biochemistry & Biophysics
DISTINGUISHED SPEAKER SERIES SEMINAR
Prof. Dmitrij Frishman, Department of Bioinformatics, Technical University of Munich, Germany.
Title: "Charting γ secretase substrates by explainable AI"
Thursday, March 23 2023, at 14:15
(Refreshments from 14:00)
Sherman building, 6th floor, Room 632, Life Sciences, TAU
Abstract: Because of its critical role in Alzheimer’s disease and cancer pathogenesis, γ-secretase is one of the most extensively studied intramembrane proteases. It cleaves the transmembrane domains of over 150 single-span membrane proteins with an extracellular N-terminus (N out). However, characteristic properties of γ-secretase substrates have remained elusive and no conserved cleavage motifs are currently known. Here we introduce a sequence-based bioinformatics method, called Comparative Physicochemical Profiling (CPP), to identify the most discriminative physicochemical features of γ secretase substrates located at specific positions within the transmembrane and juxtamembrane domains. With a balanced accuracy of 90%, our approach outperformed state-of-the-art deep protein language models (57% balanced accuracy) in predicting γ secretase substrates. Using cell-based and in-vitro assays, we validated 12 predicted substrates and 4 non-substates with a success rate of 89%. In total, we computationally classified 16.3% (n=250) of human N-out proteins as high-confidence substrates. Our approach can thus uncover subtle amino acid signatures and may have broad applicability in exploring protein sequence-to-function relationships
Host: Prof. Nir Ben-Tal, Faculty of Life Sciences, TAU
