BL: Great "what if": the kinds of experiments that can be done if computation wasn't limiting #guinformatics
3:51pm October 2nd 2014 via Hootsuite
BL: Now trying to do with splicing what they did with Myrna - ongoing work #guinformatics
3:50pm October 2nd 2014 via Hootsuite
BL: Using ReCount 2011 pub: http://t.co/g1GAv5L3pE normalizing many RNA-Seq datasets: 18 studies, 475 samples, >8B reads #guinformatics
3:49pm October 2nd 2014 via Hootsuite
BL: Next his tool Myrna, 2010 publication: http://t.co/lLUGzdshu5 eQTL calcuation - 465 at $0.34 per sample 'close to free' #guinformatics
3:48pm October 2nd 2014 via Hootsuite
BL: So the 'same SW runs on the same HW' 'and all you need is a credit card' #guinformatics
3:46pm October 2nd 2014 via Hootsuite
BL: Hadoop is open-source MapReduce, for huge clusters (like AWS). Reproducibility: a complex pipeline scales #guinformatics
3:45pm October 2nd 2014 via Hootsuite
BL: Mentions work with GEUVADIS (website http://t.co/w4F0IfgkXw ), introducing MapReduce to aggregate & compute #guinformatics
3:44pm October 2nd 2014 via Hootsuite
BL: Why study big datasets? May make new insight; may combine datasets; can add power to smaller expt #guinformatics
3:39pm October 2nd 2014 via Hootsuite
Ben Langmead (Hopkins) "Scalable software for analyzing many sequencing datasets at once" #guinformatics @benlangmead
3:36pm October 2nd 2014 via Hootsuite
IF: Costs: WES $5-30; WGS $20-100; RNA-Seq $5-10 http://t.co/K0LOYzTvM7 With: est. compute, 1mo storage, platform, support #guinformatics
3:34pm October 2nd 2014 via Hootsuite
IF:Madhavan group at GU: 78 exomes, 2TB data, 125K core-h in 1.7d All Amazon cloud-based 'at modest cost' #guinformatics
3:31pm October 2nd 2014 via Hootsuite
IF: Olopade lab: 200 exomes, 200GB data, 76K core-h in 1.25d #guinformatics
IF: Globus uses galaxy-based workflow mgmt. With Nancy Cox: 134 samples, 4TB data, 2200 core-h in 6d http://t.co/mCgC91xu87 #guinformatics
3:30pm October 2nd 2014 via Hootsuite
IF: He did early work on grid computing - 3rd-party outsourcing. Does every lab need a haystack sorting machine? #guinformatics
3:25pm October 2nd 2014 via Hootsuite
Ian Foster (U Chicago, Argonne N.L.) "Finding needles in haystacks: large-scale NGS analysis using Globus Genomics" #guinformatics
3:22pm October 2nd 2014 via Hootsuite
DK: Illustrated with chart of heart failure in-pt mortality rate by date. Area to comment, attach, similar to an intranet #guinformatics
3:16pm October 2nd 2014 via Hootsuite
(Mayo named most social-media friendly hospital: http://t.co/1C2j8kXjwb h/t @westr) #guinformatics
3:13pm October 2nd 2014 via Hootsuite
DK: Dabo, a social media platform in a healthcare environment. Getting engagement at point-of-service #guinformatics
3:12pm October 2nd 2014 via Hootsuite
Dawn Knight (Mayo Clinic, Dabo) "Making quality metrics actionable" #guinformatics
3:10pm October 2nd 2014 via Hootsuite
RT @lrasmus: @atulbutte - "the best knowledge bases today don't track if we were right or not". Great point! #GCDS2014
3:02pm October 2nd 2014 via Hootsuite
JM:Q:n of 1 trials? A:We do believe it will be multi-omics when the time comes. Schilsky (ASCO) ref: http://t.co/magNHTuH9n #guinformatics
2:43pm October 2nd 2014 via Hootsuite
JM: Putting together a center; mainly around GI cancers. "A shift from 'pay for consumption'" to "pay for outcomes" #guinformatics
2:40pm October 2nd 2014 via Hootsuite
JM: Theranostics: Two indiv's with same profile; but phosphoprotein analysis shows differences. Could be at this level #guinformatics
2:34pm October 2nd 2014 via Hootsuite
JM: 1B people w/ cancer care; "what about the rest of people worldwide?" Prospective profiling for cancer care is the future #guinformatics
2:33pm October 2nd 2014 via Hootsuite
JM: Q to audience: anyone want an allowance for medication? We are willing to pay a lot 'for meds that don't do much' #guinformatics
2:32pm October 2nd 2014 via Hootsuite
JM: Rest of world? UK uses NICE, eval of cost/quality; or cash. Almost all we have access to, NICE is disallowed. #guinformatics
JM: Innovators charge more; illust. different medications, prices, toxicity. 'The bigger Ph.III study, smaller effect' #guinformatics
2:30pm October 2nd 2014 via Hootsuite
JM: FDA approves w/o concern to cost. CMS reimburses. 'We need to have a value discussion' of what is being approved #guinformatics
2:26pm October 2nd 2014 via Hootsuite
JM: Shows EGFR diagram; 638 genes in the network from 1 receptor, for one mutated pathway. Can we measure this? #guinformatics
2:23pm October 2nd 2014 via Hootsuite
JM: Shows pathway vs. network signalling: 2nd figure messy 'like my daughter's room'. 'We just shut the door' #guinformatics
2:22pm October 2nd 2014 via Hootsuite
John Marshall: 'As oncologists we think we are giving targeted therapy', but it is just one medication for just one pathway #guinformatics
2:20pm October 2nd 2014 via Hootsuite
DH: With J&J, tranSMART platform. 2010 publication: http://t.co/4g0L1uOFtx Millenium given access to db, dev signature #guinformatics
1:51pm October 2nd 2014 via Hootsuite
Dan Houseman (Deloitte) - working on Million Veterans Project, clinical & genetic info. More collections than data gen. #guinformatics
1:48pm October 2nd 2014 via Hootsuite
Deloitte's ConvergeHealth models; one for depression shown, where he put his own demographics up; risk went from 8% to 22% #guinformatics
1:47pm October 2nd 2014 via Hootsuite
Sponsor (Deloitte) talk: Thomas Savery's 1698 steam engine patent to remove water from coal mines. http://t.co/cmxyQgpmtD #guinformatics
1:45pm October 2nd 2014 via Hootsuite
.@PinkyG123 After ID of their 53 gene signature, subsequent samples done via NanoString. See their publication here: http://t.co/zQLPUxCmJu
1:17pm October 2nd 2014 via Hootsuite in reply to
.@atulbutte At today's #guinformatics talk you certainly emerged a rockstar. (First time I've heard you speak, an inspiring talk!)
12:12pm October 2nd 2014 via Hootsuite in reply to
KN: Shows complex flowchart - to account for sample sources, differences in dosage and other variables #guinformatics
12:11pm October 2nd 2014 via Hootsuite
KN: From only 113 samples: Manhattan plot shows interesting results but not passing significance test; larger #'s planned #guinformatics
12:10pm October 2nd 2014 via Hootsuite
KN: Review of genetic component of auto-immune diseases; PGx have a lack of selection, so large odds ratios, few samples #guinformatics
12:06pm October 2nd 2014 via Hootsuite
KN: Studied for many years; 60% of 2901 pts had immune related AE: lower GI, others. Looking for underlying variation #guinformatics
12:03pm October 2nd 2014 via Hootsuite
KN: Two immunotherapies are anti-CTLA4 in '11, anti-PD1 a few weeks ago.Tcell/APC blocking CTLA4 (ipilimumab) #guinformatics
12:01pm October 2nd 2014 via Hootsuite
KN: Geneticist by training; review of melanoma treatments in '11 for somatic BRAF mutations (~45% of pts). 2 immunotherapies #guinformatics
11:59am October 2nd 2014 via Hootsuite
Kathernine Nathanson (UPenn) "Germline predictive biomarkers of immune response and toxicity" #guinformatics
11:58am October 2nd 2014 via Hootsuite
JG:Q:How to discriminate predictive vs. prognostic? A:'all ongoing' b/c the tool has a lot of potential, new area #guinformatics
11:57am October 2nd 2014 via Hootsuite
JG: Referenced this paper on CD40 (next speaker) that came out last week for melanoma immunotherapy http://t.co/zklsx8tgeB #guinformatics
11:54am October 2nd 2014 via Hootsuite
JG: % of T-cells vs. clonality: those with high T-cells, high clonality were most likely to respond to anti-PD1 #guinformatics
11:51am October 2nd 2014 via Hootsuite
JG: Collab with UCLA Tony Ribas CTLA4 blockade data shown. Another: anti-PD1 therapy (unpubl., manuscript submitted) #guinformatics
JG: QuanTILfy(tm): prognosis for accurate staging based on count/clonality; work on standardization in progress #guinformatics
11:49am October 2nd 2014 via Hootsuite
JG: Also tumor immunol.; get TIL prognosis and clonality. Quantilfy / TIL-seq (same name); future app: transplantation #guinformatics
11:48am October 2nd 2014 via Hootsuite
