October 2015

Durbin #icg10 Next: Raw data to the entire dataset? Take TGP raw reads? A full text index of all reads. 87 TB of data!

12:01am October 23rd 2015 via Hootsuite

Durbin #icg10 Haplotype searches have been linear to now. Now indexed version is 1000x faster. Shows HRC.I imputation data after compression

11:59pm October 22nd 2015 via Hootsuite

Durbin #icg10 gzip is a factor of 5.9x But scales to 133x fold at 100k sequences. Take entire HRC, 32k samples and 39M SNPs in only 6.1 GB

11:57pm October 22nd 2015 via Hootsuite

Durbin #icg10 Variant of BWA, positional BWA, a new data structure for haplotype matching. Sorting in BWA and blocks compressed

11:55pm October 22nd 2015 via Hootsuite

Durbin #icg10 HRC has 35K samples, and an Imputation Server setup at UMich and Sanger.

11:53pm October 22nd 2015 via Hootsuite

Durbin #icg10 Reviews imputation and its current ability to handle 1,000’s. Need to scale to millions. Haplotype Ref Consortium

11:53pm October 22nd 2015 via Hootsuite

Durbin #icg10 Can use for LD-based imputation, statistical phasing of new samples. Error-correct low-coverage WES/WGS data

11:50pm October 22nd 2015 via Hootsuite

Durbin #icg10 HapLotype reference panels: TGP up to 26 pop’s; 5,008 haplotypes. HRC.I has 65K haplotypes from 15 EU pop’s 39.2M var’s

11:49pm October 22nd 2015 via Hootsuite

Durbin #icg10 Can map a 39x genome in 40h. Next: extend BWA to graphs called BWT, min unique k-mets w/one base prefix ext. 20GB index

11:47pm October 22nd 2015 via Hootsuite

Durbin #icg10 New software ‘vg’ on Github, edges and nodes. Every 10bp has a branch-point; 100’s of M of var’s

11:45pm October 22nd 2015 via Hootsuite

Durbin #icg10 Map to a structure incl. known var; >99% var’s/person seen before. A variation graph. Many genomes in one non-redundant str

11:43pm October 22nd 2015 via Hootsuite

Durbin #icg10 BWA-MEM now handles var’s correctly, ALTs and decoys. But go beyond: would like to build unk var into the reference

11:42pm October 22nd 2015 via Hootsuite

Durbin #icg10 GRC: NCBI, EBI, Sanger, WashU: fix errors, fill gaps. GRCh38 Dec 13, fixes >7K var’s in TGP. 261 Alt loci in 192 regions.

11:40pm October 22nd 2015 via Hootsuite

Durbin #icg10 Complexity of the reference and novelty; bias to what we’ve seen vs unbiased ref-free assembly.

11:37pm October 22nd 2015 via Hootsuite

Durbin #icg10 Seq acquisition increases 2x-4x/y; need Fast, Accurate methods on Compressed data. Keep finding variation seen before.

11:36pm October 22nd 2015 via Hootsuite

Durbin #icg10 The HGP was a reference; many uses illustrated. Major problem is alignment errors. A critical area. https://t.co/LUcKiGSOYH

11:34pm October 22nd 2015 via Hootsuite

Richard Durbin (Sanger Inst) #icg10 Building and using new genome reference structures

11:30pm October 22nd 2015 via Hootsuite

Olson #icg10 Me-too drugs, over-rely on biologicals, unrealistic insistence ‘absolute’ drug safety, over-Dx and over-Rx of pre-sympt. indivs

11:28pm October 22nd 2015 via Hootsuite

Olson #icg10 And the middle class is under-medicated, and can fuel a research effort. Things not to do: following the US/EU model

11:25pm October 22nd 2015 via Hootsuite

Olson #icg10 An China is the’obvious place’ - ‘Well-poised to dominate’ and it may take 50y. Potential to disrupt: huge market...

11:24pm October 22nd 2015 via Hootsuite

Olson #icg10 Transition to drug dev: thinks the big opportunities are charting path toward better Rx - so many therapies are ineffective.

11:21pm October 22nd 2015 via Hootsuite

Olson #icg10 Susan Hellman’s editorial, Toward PM ‘a new social contract’ - explains, sensitive to Chinese audience what it may mean

11:20pm October 22nd 2015 via Hootsuite

Olson #icg10 Inherently reciprocal - what pts. will gain. PLOS Biol PMID 25369215 Redefining Genomic Privacy: trust and empowerment.

11:18pm October 22nd 2015 via Hootsuite

Olson #icg10 A challenge: building trust w/individual participants. Much higher transparency is needed than what is done now. Pt. control.

11:15pm October 22nd 2015 via Hootsuite

Olson #icg10 The HGP: a player’s perspective PMID 12079320 2002 J Mol Biol ref

11:14pm October 22nd 2015 via Hootsuite

Olson #icg10 The HGP ‘was hard-won’; it could have turned out differently. 2002 paper JMB ‘A Player’s Perspective’ still a useful read.

11:10pm October 22nd 2015 via Hootsuite

Olson #icg10 Int’l Congress of Genetics in 1988: got substantive Chinese participation. Scientists share responsibility in the HGP

11:09pm October 22nd 2015 via Hootsuite

Olson #icg10 NRC’s 1988 report on seq and mapping the human genome, and the 2011 Toward Precision Medicine: he served on both committees.

11:07pm October 22nd 2015 via Hootsuite

Maynard Olson (Univ Wash) #icg10 How genomics contribute to improved treatments https://t.co/oT2yVrRnbP

11:05pm October 22nd 2015 via Hootsuite

Paranjape #icg10 Promose of clinicians and researchers together making Precision Med possible - make ‘Omics affordable, accessible

10:14pm October 22nd 2015 via Hootsuite

Paranjape #icg10 Anticipates BGISEQ-500 will have dozens of apps (incl ChIP-Seq), and is done locally or on cloud w/ID interface

10:13pm October 22nd 2015 via Hootsuite

Paranjape #icg10 Will be open-source on Github, and connect across public datasets. Extend to add’l partners. Working w/Aliyu, BGI

10:11pm October 22nd 2015 via Hootsuite

Paranjape #icg10 Intel’s Collaborative Cancer Cloud, launches early 2016; includes image phenotype and pt history

10:09pm October 22nd 2015 via Hootsuite

Paranjape #icg10 Woks w/OHSU and others: wants to move analytics to the data; do it in an open secure trust model, and at any scale

10:07pm October 22nd 2015 via Hootsuite

Paranjape #icg10 States 96% of data is not available, too large to move, and protected by privacy laws and commercial concerns.

10:05pm October 22nd 2015 via Hootsuite

Paranjape #icg10 Data size large, Dx and Rx takes weeks, privacy protection hard, scale of training a huge problem (w/o expertise it breaks)

10:04pm October 22nd 2015 via Hootsuite

Paranjape #icg10 3M pts in China is >7 Exabytes of WGS data. Goal: reduce data generation & analysis from months to 1 day by 2020.

10:01pm October 22nd 2015 via Hootsuite

Kenyan Paranjape #icg10 (Intel) Scaling Precision Health. The challenge of an aging population. https://t.co/OLVKVvzJIi

9:56pm October 22nd 2015 via Hootsuite

Jun Wang #icg10 brings up the potential of @icahninstitute Resilience Project - genomic outlier ‘Superhero’ (although not by name, alas).

9:52pm October 22nd 2015 via Hootsuite

Jun Wang #icg10 “I like ‘patients like me’, but it needs to become ‘people like me’”

9:50pm October 22nd 2015 via Hootsuite

Jun Wang at #icg10 hops onto single cell WGS in two GBM pts, and clonal evolution. ‘Probably for cancer you have to do it’

9:46pm October 22nd 2015 via Hootsuite

Jun Wang at #ICG10 presents AI-guided breeding (machine learning genotype to phenotype) with controlled env breeding of Foxtail millet.

9:44pm October 22nd 2015 via Hootsuite

Jun Wang #icg10 believes after 1M WGS the value to the individual will rise from not $100 today, but over $10K then.

9:38pm October 22nd 2015 via Hootsuite

Jun Wang at #icg10 emphasizes importance of LFR for phasing, the HTP Revolocity, orig wanted 1M/yr, not 10K.

9:36pm October 22nd 2015 via Hootsuite

Jun Wang at #ICG10 hopes to offer free WGS to everybody ‘soon’; extrapolates $1 WGS in 2019. Value to individual is not $100 now...

9:35pm October 22nd 2015 via Hootsuite

Jun Wang makes the case for more data at #icg10 - Million Omics Database from individuals. https://t.co/bhPLHey3Ky

9:33pm October 22nd 2015 via Hootsuite

Jun Wang claims his own genome may be the most accurate (incl. phasing) at #icg10 (better than GIAB NA12878?) due to multi-platform NGS

9:28pm October 22nd 2015 via Hootsuite