Learning to Untangle Genome Assembly Graphs - IPAM at UCLA

Explore a groundbreaking approach to genome assembly using deep learning in this 43-minute lecture by Xavier Bresson from the National University of Singapore. Delve into the challenges of untangling genome assembly graphs and discover how graph convolutional networks can be trained to resolve these complex structures. Learn about the innovative framework that outperforms traditional hand-crafted heuristics in reconstructing chromosomes, achieving higher accuracy and improved assessment metrics. Gain insights into the experimental setup, dataset generation, and evaluation methods used to demonstrate the model's remarkable ability to generalize across different chromosomes.

Intro
Outline
Genome Sequencing Machine
Genome Assembly Problem
Challenges
Raven Genome Assembler
Machine Learning Framework
Assembly Graph Construction
Edge Labeling
Designing GNNs for Assembly Graphs
Edge Prediction Layer
Network Training
Experimental Setting
Evaluation
Dataset and Code