Molecular Characterization for Diagnoses and Treatment
summary
Proposes a genome graph reference that stores common variants (roughly ≥1%) as DNA-string nodes with bi‑directed joins; builders used CACTUS, de Bruijn and the VG mapper to test BRCA1/2, MHC, KIR and SMA. Results: k‑mer recall on MHC ~99%, unique and perfect read mapping rises from ~80% to ~90% versus the linear reference, and variant calling on platinum genomes outperforms the GATK pipeline. Augmentations needed drop dramatically (primary MHC needs ~4,500; rich graphs need far fewer). Also tested a 1000 Genomes baked graph.
AI-written summary · about
chapters (22)
- 0:00Intro
- 0:19Third example: building the universal map of human genome variation
- 3:33Human Genome Variation Map
- 5:04At base level we assign unique position identifiers
- 6:39Unique identifiers stay the same as the reference grows
- 7:39Talk Overview
- 7:49Genome Graphs
- 11:12Under the Hood at GA4GH
- 13:05HGVM Pilot Bake-Off
- 13:45Graph Submissions
- 14:21Diversity of approaches/input data
- 17:10Graph Examples (BRCA2)
- 22:22Basic Stats
- 23:56Average Cut Width
- 27:381000 Genomes Project k-mer overlap
- 29:22Read-mapping
- 32:14VG Variant Calling Pipeline
- 35:32VG Call Details
- 39:54Illumina Platinum Genomes Comparison
- 40:39Genome Graph Comparison
- 42:08Topology Comparison
- 44:48Summary
related talks
30mUnlocking the Genome: Exploring the Potential of CRISPR/Cas9
12mxBGAS: A Bridge Proposal For RV128 And HPC
49mKeynote Address: Everything You Always Wanted to Know about Storage Analysis...
27mComputers for Super-Human AI
01h 22mTutorial 5: The GCT Chasm II
26mHere's to History: Programming Through Archaeology
14mFoundational HPC Systems for 2020 and Beyond
31mAnswering FAQs in CSPs, PGMs, Databases, Logic and Matrix Operations