Roll over the wheel to learn more about how we apply artificial intelligence to three different interfaces in the immune synapse.
PROPRIETERY ANTIGEN-MHC BINDING MODEL
Decreases library sizes by magnitudes, by efficient prediction of which epitopes could present on each MHC allele based on a combination of experimental data on peptide presentation and theoretical definition of pertinent features to explain binding relationships. The model already surpass publicly available algorithms such as netMHC and is continuously updated and trained with experimental data such as folding potential of produced tetramers.
TCR SEQUENCE PROBABILITY
Theory and date driven predictor of T cell sequence generation probability to enable sequencing data interpretation, disease-related TCR selection, and efficient exploration of TCR sequence space for predictive and generative modeling.
T CELL PHENOTYPE DESCRIPTION
Initial modeling to predict current state and future trajectories of an immune cell base on select combinations of gene-expression profiles and surface markers to identify clones based on current cytotoxic and exhaustion status and future potential.
TCR CROSS REACTIVITY ASSESMENT
Description of 3D TCR-peptide-MHC structure using adjacency matrix of pairwise distances. These extremely high-throughput predicted structures can be evaluated by similarity indexes and structure-based affinity prediction models to asses potential cross-reactivity.
AUTOMATIC HIT IDENTIFICATION
Hit browser automatically counts individual tetramer molecules bound to single cells, quantity enrichment of binding within clonotypes relative to controls, identifies and summarizes hits within and across samples for long-range and interactive data analysis
ANTIGEN-MHC-TCR STRUCTURE BINDING PREDICTOR
Using deep learning to combine structural simulations of pMHC-TCR interactions with Cogen’s platform data yields a powerful new approach to decode the immune system. First step towards shifting experimental paradigm from exploratory to confirmatory through predictive and generative modeling.