Debugging Genomic Profiling Experiments and Predictive Models with Interpretation Tools

Explore a comprehensive lecture on debugging genomic profiling experiments and predictive models using interpretation tools. Delve into the world of deep learning models in genomics as Anshul Kundaje from Stanford University demonstrates how interpretation frameworks can identify latent confounders, technical biases, and artifacts in experimental data. Learn about designing models that accurately capture and correct technical biases, resulting in high-performance predictive models that reveal novel insights into gene regulation mechanisms. Discover how these efficient and robust models outperform state-of-the-art foundation models while providing near-optimal latent bias correction. This talk, part of the AI≡Science series, strengthens the bond between artificial intelligence and sciences, offering valuable insights for researchers and practitioners in the field of genomics and AI.

Debugging genomic profiling experiments and predictive models with interpretation tools