Fast and Scalable Inference of Cancer Cell Lineages Using Multi-sample Deep Sequencing Somatic SNVs
summary
A tool called litchi infers tumor lineage trees from deep multisample somatic SNV frequencies by grouping variants into binary presence profiles, GMM clustering by frequency, and building an evolutionary constraint network whose valid spanning trees satisfy the perfect phylogeny sum and subset constraints. A min set cover heuristic and optional user-provided presence calls handle ambiguous low-frequency calls; it runs fast, produces the same trees as published analyses for most cases on a dataset of 587 validated somatic SNVs across multiple biopsies from eight patients, and includes a GUI and GMM clustering hooks.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:15Intra-Tumor Heterogeneity
- 0:33Multi-Sample Sequencing Studies
- 2:16Input Data
- 5:02Perfect Phylogeny Model Assumption
- 5:28Perfect Phylogeny Model: Constraints
- 7:10LICHEE's Problem Formulation
- 10:35LICHE: Method Overview
- 11:16Presence Patterns Across Samples
- 13:21VAF-Based Clustering
- 14:06Evolutionary Constraint Network
- 14:54Lineage Tree Search
- 15:25CCRCC Study by Gerlinger et. al (2014)
- 18:44LICHEE Runtime DEMO Movie
related talks
33mEvolution of Music, Cantometrics
30mUnlocking the Genome: Exploring the Potential of CRISPR/Cas9
21mEnabling the next level of scientific exploration
26mTed Willke, Senior Principal Engineer, Intel Labs
01h 35mThe Twin Algorithms
58mStrong Generalization from Small Brains and No Training Data
31mAnswering FAQs in CSPs, PGMs, Databases, Logic and Matrix Operations
43mThe Brain’s Guide to Dealing with Context in Language Understanding