(Just realized that for the past 4 days 1am is rolling in early, the #AGBT14 Marriott lobby confab et al)
8:04am February 16th 2014 via Hootsuite
Coming home from DCA, my tire picks up a nail on the GW Pkwy. I roll in about 1am, and pretend that 'it's just another day at #AGBT14'
7:55am February 16th 2014 via Hootsuite
.@GroovyGeek Sorry about the borked link. Here you go: http://t.co/cvOrLzAWYh
1:05am February 16th 2014 via Hootsuite in reply to
Here's a fixed link to the Next Generation Technologist blog post: #AGBT14 thoughts and Jeffrey Schloss' Talk http://t.co/mZU8Ousqzm
1:05am February 16th 2014 via Hootsuite
.@BekahMcLaughlin Argh. I'm unfollowing @cathcoste ASAP, Sorry! #AGBT14
9:28pm February 15th 2014 via Hootsuite in reply to
New post from the Next Generation Technologist: Post #AGBT14 thoughts and Jeffrey Schloss' Talk http://t.co/ZytHdb4Txv
9:16pm February 15th 2014 via Hootsuite
RT @cathcoste: Gilbert's #AGBT14 talk was funny even on Twitter; see Storify. http://t.co/sFDgOWrlHY @DaleYuzuki @lexnederbragt @neilhall_uk
6:15pm February 15th 2014 via Hootsuite
Porreca: Workflow - the balance between quality and cost of assay. 'More steps = more places to fail' #AGBT
3:16pm February 15th 2014 via Hootsuite
Porreca: In silico validation - can introduce synthetic reads at a certain frequency to test your analysis sensitivity #AGBT
Porreca: Looking at indels, can be very difficult to call. #AGBT
3:15pm February 15th 2014 via Hootsuite
Porreca: Contam. of barcodes - just from a 96-well plate lid being used as a flipped orientation. Trace contam., but detectable #AGBT
3:12pm February 15th 2014 via Hootsuite
Porreca: Barcoding QC - monitoring stock reagents to insure no contamination effects. Does synthetic target sequencing to det error #AGBT14
3:11pm February 15th 2014 via Hootsuite
Porreca: MIP diagrammed, showing data from 2014 Genet Med ref here. http://t.co/mwTfM3B6r1 #AGBT14
3:10pm February 15th 2014 via Hootsuite
Porreca: Did Sanger comparison to their calls; ~7M TN, 4K TP, 8 FP, 1 FN. (Sample had an aneuploidy event, clear upon inspection) #AGBT14
3:08pm February 15th 2014 via Hootsuite
Porreca: Can add to cost / complexity if capture technology isn't always consistent #AGBT14
3:07pm February 15th 2014 via Hootsuite
Porreca: Focus is on accuracy, workflow, and integration. 'Completeness': you have to interogate every site. For CF = all ~550 muts. #AGBT14
3:06pm February 15th 2014 via Hootsuite
Porreca: They use MIP-based target capture for complexity reduction. Tuned to ID different classes of disease-causing muts #AGBT14
3:05pm February 15th 2014 via Hootsuite
Porreca: Uses NGS and other complementary tech for Fragile X, Alpha Thalassemia etc. at Good Start Genetics #AGBT14
3:04pm February 15th 2014 via Hootsuite
Porreca: Accuracy / sensitivity in carrier screening is sensitive to FN's - both partners need positive result to have risk. #AGBT14
3:03pm February 15th 2014 via Hootsuite
MacArthur: For 26K joint-called ones: ave 67 unique SNPs for Mendelian disease pts. Major factor: consistent var calling #AGBT14
2:50pm February 15th 2014 via Hootsuite
MacArthur: NHLBI ESP samples n=6.8k - cp to ave. Mendelian dis. pt., ave 569 unique SNPs. #AGBT14
MacArthur: No can assign a quantitative score to variants in Mendelian patients. #AGBT14
2:49pm February 15th 2014 via Hootsuite
MacArthur: Comparing obs vs. predicted syn muts, missense, LoF variants. Genes ID'd subject to human constraints. LoF depletion #AGBT14
2:48pm February 15th 2014 via Hootsuite
MacArthur: Joint calling smooths batch-to-batch variation; plots Var count / kb; and minor allele count, across 40K exomes #AGBT14
2:45pm February 15th 2014 via Hootsuite
MacArthur: Sort by SNP count by ancestry across 40K - naturally Africans had much higher counts #AGBT14
2:42pm February 15th 2014 via Hootsuite
MacArthur: Raw BAM are ~1PB. Reduced ~1TB, 74K CPU h, 84K h for var calls; some regions req 64GB RAM. VCF's final = 2.35TB. (Whew!) #AGBT14
MacArthur: How 2.6K samples are filtered out. Relatedness of ethnicity plot via geography; strong conc. of European samples. #AGBT14
2:41pm February 15th 2014 via Hootsuite
MacArthur: From 26K pilot to 57K production project finished 2w ago. Of this 40K filtered reported on now (prelim). #AGBT14
2:39pm February 15th 2014 via Hootsuite
MacArthur: Het compression can reduce complexity, just a few synthetic reads. 10x-20x BAM compression; GATK UnifiedGenotyper scales #AGBT14
2:38pm February 15th 2014 via Hootsuite
MacArthur: GATK working on reduced BAMs. IGV shot: Homozygotes can collapse down to a single synthetic meta-read; keep the others #AGBT14
2:37pm February 15th 2014 via Hootsuite
MacArthur: "Can we simultaneously call variants across 10's of thousands of exomes?" 10GB/exome, 'not easily' #AGBT14
2:36pm February 15th 2014 via Hootsuite
MacArthur: 'Technical challenges are substantial' to look at large-scale exome sequencing. Illus. joint calling, FP and FN errors #AGBT14
2:34pm February 15th 2014 via Hootsuite
Boy am I glad I stayed for Hakonarson's great talk. Tired but tremendous work being done at CHOP. Here's to #AGBT15. #AGBT14
1:58pm February 15th 2014 via Hootsuite
Hakonarson: 'Oh, and a 100 thousand more things': 100K WGs put through IVA and 12y phenotype data, to understand common disease #AGBT14
1:56pm February 15th 2014 via Hootsuite
Hakonarson: NOTCH3 inhib. for myofibramatosis, can put together a PGx model in rare disease. Can go from rare disease to fast-track #AGBT14
Hakonarson: 50% of 49 genes found are novel; many of therapeutic interest with repositioning oppy's. NOTCH and many others #AGBT14
1:54pm February 15th 2014 via Hootsuite
Hakonarson: Diagnostic odyssey paper here in Genome Med. http://t.co/Kwvbj2gSBr #AGBT14
1:53pm February 15th 2014 via Hootsuite
Hakonarson: Everything through IVA, 280 rare dis families, 49 genes so far. List of several new genes. A 20-y diag. odyssey solved #AGBT14
1:52pm February 15th 2014 via Hootsuite
Hakonarson: They use AMZN for everything; <14h / exome, laid out their pipeline. 10GB seq, 12GB align, 100MB vars (not bases, bytes) #AGB
1:51pm February 15th 2014 via Hootsuite
Hakonarson: Also rare disease prog; 525 families, 1598 indiv's. 49 disease genes to-date 'just scratching surface' of samples onhand #AGBT14
1:50pm February 15th 2014 via Hootsuite
Hakonarson: Medication repositioning of a therapy - based on molecular phenotyping. Neuroblastoma ALK #AGBT14
1:49pm February 15th 2014 via Hootsuite
Hakonarson: ADHD - est $76B/y cost in US; most common disease among children; discovered var's Nature Gen '11 http://t.co/4UHoqVRDoG #AGBT14
1:48pm February 15th 2014 via Hootsuite
Hakonarson: And neuroimaging to boot; interview-based dis. class.; top and bottom 15% of the population (extreme phenotypes) #AGBT14
1:46pm February 15th 2014 via Hootsuite
Hakonarson: Power of the biobank - 10K recruited back for add'l phenotyping, assessment, 'computerized neurocognitive battery' #AGBT14
1:45pm February 15th 2014 via Hootsuite
Hakonarson: Biobank a key asset: EHR has unlimited potential - longitudinal followup, history of medications, track complex pheno's #AGBT14
1:44pm February 15th 2014 via Hootsuite
Hakonarson: Focus on neurodevelopmental studies, other developmental ones. History: Linkage, then GWAS, then WES. #AGBT14
1:43pm February 15th 2014 via Hootsuite
Hakonarson: CHOP Center for Applied Genomics - 2006 founded; 75 staff; 60K children recruited. Target: 100K. EHR goes 12y back #AGBT14
1:42pm February 15th 2014 via Hootsuite
Hakonarson: Where medicine is headed: stem-cell therapy has major promise, will take time for acceptance. 'Opportunities are huge' #AGBT14
1:41pm February 15th 2014 via Hootsuite
Hakonarson: Disclosing as a founder of neuroFix Therapeutics. Chronic diseases are growing. 15% of children (10M) US have spec needs #AGBT14
1:40pm February 15th 2014 via Hootsuite
Up next: Hakon Hakonarson (CHOP): Impact of optimized seq analysis pipeline on decision making in genomic medicine #AGBT14
1:38pm February 15th 2014 via Hootsuite
