Gamgee: A C++14 library for genomics data processing and analysis
summary
Broad genomics pipeline is reimplemented in a C++14 library to remove Java memory indirection, use contiguous byte arrays, shared_ptr-backed views and builder-style mutation, and exploit SSE/AVX and GPUs. Joint genotyping originally costs ~7.6 CPU days per genome; a C++ port yields 9× faster on CPU and up to 154× on a Tesla K40 with CUDA. File I/O and in-memory layout speed BAM/VCF reads ~17×, mark-duplicates runtime drops from ~11 hours to ~1 hour per genome and exomes to minutes. Gamgee C++14 library integrates with existing GATK JNI bridge.
AI-written summary · about
chapters (11)
- 0:00Intro
- 1:33To fully understand one genome we need
- 4:00Improving human health in 5 easy steps
- 16:00Identifying mutations in a genome is a
- 21:48Unfortunately, real data
- 27:06Data dependencies of each cell in
- 28:57Heterogeneous compute speeds
- 49:45Writing tools to perform operations
- 53:23A lambda configurable class
- 54:38Coverage distribution tool: functional style
- 55:19The future of the GATK
related talks
46mScalable Genomics with ADAM
17mBig data in personalized medicine
25mScaling Up Genomics with Spark
53mApplied Functional Programming in Computational Genomics
21mInteractive Machine Learning on Genomics Data with Spark Notebook
27mA Statistical, Reference-Free Algorithm Subsumes Myriad Problems in Genome Science
50mCancer Genomics - a biologist and a developer
51mSimplicity For Scale- Analyzing 15 Million DNA Samples With Python