Getz: Know there are more genes to be found; looked at 5K cancer genomes from 21 tumor types, Lawrence 2014: http://t.co/Oosoqya5C3 #AACR14
8:58pm April 8th 2014 via Hootsuite
Getz: Charts from several TCGA papers - genes mutated more often than by chance; the long tail ends b/c lack stat power #AACR14
8:56pm April 8th 2014 via Hootsuite
Next: Gad Getz, Broad Inst. Why do we need a cancer data and knowledge commons? A place to store & share; need to harmonize results #AA
8:54pm April 8th 2014 via Hootsuite
Haussler: 2 months of YouTube is 100PB. 'We can do this." "We must head off the dev. of genomic information silos" #AACR14
8:47pm April 8th 2014 via Hootsuite
Haussler: Need machine-aided ways for pattern recognition: gather data, preprocess, train + eval., store + display, then predict #AACR14
8:46pm April 8th 2014 via Hootsuite
.@obigriffith Whoops - too fast on the <ctrl>-c. (And a looong day - I'm just about 'done'!)
8:44pm April 8th 2014 via Hootsuite in reply to
Hudson:Venn diagram of same BAM, different centers calling var's. Link to UC Berkeley's SMaSH website http://t.co/dTincSvLbK #AACR14
8:42pm April 8th 2014 via Hootsuite
Hudson: But cancer is the 'high water mark for genomics' due to its complexity; other diseases cost less for data stor & analysis #AACR1
8:39pm April 8th 2014 via Hootsuite
Hudson: In 2014, est cost is $50/genome/yr for 1M genomes (~100PB); includes 25K disks plus 100K cores incl. operating costs #AACR14
8:38pm April 8th 2014 via Hootsuite
Hudson: Not monolithic geographically, nor architecture: read-layer db (100PB); variation layer db (1PB); interpret layer (1TB) #AACR14
8:37pm April 8th 2014 via Hootsuite
Hudson: Individual, uniform consent process for 1M genomes; 'it's API's, not file formats' "we have to get beyond that" #AACR14
8:34pm April 8th 2014 via Hootsuite
Haussler: Computation needs to be provided close to the data. Dave Patterson's '1M genomes warehouse' link http://t.co/G3l71KD2Uv #AACR14
8:33pm April 8th 2014 via Hootsuite
Haussler: A task team is working on development a next-gen human genetic variation including known variation + API to access it #AACR14
8:29pm April 8th 2014 via Hootsuite
Haussler: Need API's and computational procedures on top of that (where the innovation occurs). But API standards needed first #AACR14
8:27pm April 8th 2014 via Hootsuite
Haussler: File formats - BAM, CRAM and VCF's. But inevitably - will need to add future enhancements for clinical applicability #AACR14
8:26pm April 8th 2014 via Hootsuite
Haussler: Example of a driver project: "Genomic Matchmaker" - Heidi Rehm, expert sharing of rare genetic variants #AACR14
8:25pm April 8th 2014 via Hootsuite
Haussler: Founding partners 70 institutes, 40 countries; several working groups; also several driving projects #AACR14
8:24pm April 8th 2014 via Hootsuite
Haussler: 'Major medical centers... create Balkanized, incompatible, inadequate systems'; 'reinforces barriers to data sharing' #AACR14
8:22pm April 8th 2014 via Hootsuite
Haussler: The future needs multiple CGHubs worldwide - major cloud providers (AMZN, GOOG, MISFT) should be engaged #AACR14
8:21pm April 8th 2014 via Hootsuite
Haussler: Used commodity HW, now has served >1M total files; 15PB transferred; has 1.4PB data; 4 Gb/s bandwidth (!) peak 15 Gb/s #AACR14
Haussler: CGHub is ~$100/yr/genome at 50K genomes; growing to 2.5PB in near future; scalable to 20PB (20 x10^15 bytes) #AACR14
8:19pm April 8th 2014 via Hootsuite
Haussler: TCGA is scaling to 10K tumors from 20 adult cancers; ICGC will hit 25K tumors; Cancer Genomics Hub in San Diego CA #AACR14
8:18pm April 8th 2014 via Hootsuite
Next: David Haussler, UCSC. One key difference between HIV and cancer: it isn't passed onto cancers of our children #AACR14
8:17pm April 8th 2014 via Hootsuite
Hudson: Modeled after the W3C, 12 on the transitional steering committee. List of steering committee: http://t.co/MJhXkYg5ri #AACR14
8:15pm April 8th 2014 via Hootsuite
Hudson: Global Alliance: a startup began in Jan 2013; white paper Link to UCSF (PDF) http://t.co/f9WErjIh6n #AACR14
8:12pm April 8th 2014 via Hootsuite
Hudson: ICGC datasets too large to handle due to sheer size; want a commons database to query across datasets #AACR14
8:10pm April 8th 2014 via Hootsuite
Hudson: ICGC has standardized datasets. Portal launched Oct 2013, homepage here: http://t.co/V2RSl7ZRX9 #AACR14
8:09pm April 8th 2014 via Hootsuite
Forum: How to Achieve a Cancer Knowledge Commons Database of Cancer Genetics, moderator Tom Hudson, OICR
8:07pm April 8th 2014 via Hootsuite
Bass: Conclude TCGA offers 'numerous opportunities for further detailed mechanistic and translational studies' #AACR14
7:57pm April 8th 2014 via Hootsuite
Bass: ID gene RHOA plays role in invasion, highly recurrent mutations. Tyr42 - RHOA is a GTP-ase protein, like Ras; RHOA + Rock1 #AACR14
7:56pm April 8th 2014 via Hootsuite
Bass: Moving onto Stable type: genomically-stable. Used Mut-Sig tool to find mutations higher than chance #AACR14
7:54pm April 8th 2014 via Hootsuite
Bass: Greet et al Hum Cancer Biol 2012 ref. http://t.co/lptslLeIKj confirms similar effect of EBV+ in lymphomas inducing PD-L1 #AACR14
7:52pm April 8th 2014 via Hootsuite
Bass: TCGA-enabled wide analysis (esp. methylation data) helped define EBV+ group; CNV found 9p ampl; exp data showed PD-L1/L2 #AACR14
7:50pm April 8th 2014 via Hootsuite
Bass: Strong T-cell infiltration in EBV+ gastric ca: expression signature of enriched signalling. #AACR14
7:49pm April 8th 2014 via Hootsuite
Bass: PDL1/2: immune checkpoint markers. Found EBV+ gastric ca had much higher expression of PD-L1/L2. #AACR14
7:48pm April 8th 2014 via Hootsuite
Bass: For EBV+ gastric ca, found focal amplification at 9p24.1, JAK2 locus. Other genes nearby - DC274 & PDCLILG2 (PDL1 and PDL2!) #AACR
7:47pm April 8th 2014 via Hootsuite
Bass: Found dramatic rates of PIK3CA mutation in EBV+ gastric ca - on the order of 80% #AACR14
7:45pm April 8th 2014 via Hootsuite
Bass: EBV+ 'King' of hypermethylation (of all types of cancers TCGA has analyzed to-date) #AACR14
Bass: Looking at methylation profiling for EBV+ and MSI+ gastric ca: across 1315 CpG sites, clear differences #AACR14
7:44pm April 8th 2014 via Hootsuite
Bass: Four types: EBV, MSI, 'Stable', Aneuploid. Stable - mainly diffuse. EBV - mostly fundus body #AACR14
7:42pm April 8th 2014 via Hootsuite
Bass: Four groups: eBV #AACR14
7:41pm April 8th 2014 via Hootsuite
Bass: Cp to other TCGA, gastric took it from an unbiased classification approach. Det 4 subtypes via unsupervised clustering #AACR14
Bass: TCGA 295 gastric ca's - including GEJ. Goal: better classification; ID key pathways; ID targets/biomarkers #AACR14
7:39pm April 8th 2014 via Hootsuite
Bass: For gastric / esophageal adenocarcinoma - histological, anatomic, geographic, molecular differences. #AACR14
Bass: Reviewed types of somatic alterations: mutations, amplifications, deletions, translocations #AACR14
7:37pm April 8th 2014 via Hootsuite
Next: Adam Bass, Dana-Farber Cancer Inst. "Emerging insights from the TCGA study of gastric cancer" #AACR14
7:34pm April 8th 2014 via Hootsuite
Fox: Conclude: intestinal helminths can attenuate gastric disease. (Showed some data from Columbia, South Am.) #AACR14
7:32pm April 8th 2014 via Hootsuite
,@splon That is quite a coincidence indeed!
7:30pm April 8th 2014 via Hootsuite in reply to
Fox: Now in INS-GAS-FVP mice, now with H. polygyrus: reduced dysplasia, increase of FOXP3 regulatory cells #AACR14
7:27pm April 8th 2014 via Hootsuite
Fox: Looking at helminths (Heligmosomoides polygyrus), a natural mouse parasite and its effect with H. felis alone #AACR14
7:26pm April 8th 2014 via Hootsuite
