Q: Parallel review? Denham: If you are going for nat'l coverage, if the answer is 'no' 'you are done'. Putcha: no-one to rescue you #NGDx16
9:57am August 23rd 2016 via Hootsuite
Putcha: FDA approval does not mean payer reimbursement; until they evaluate utility claims in label (ie Exact Sci) #NGDx16
9:54am August 23rd 2016 via Hootsuite
Putcha (MolDx, Palmetto) Want to encourage datasharing, coverage w/evidence development (data collection). Not commonly seen #NGDx16
9:52am August 23rd 2016 via Hootsuite
Denham: Context impt Hilborne (UCLA, Rand): Has to be part of the study, asking Q's up-front. To ask 'how will it make difference?' #NGDx16
9:50am August 23rd 2016 via Hootsuite
Reimbursement at #NGDx16 - what about data-sharing w/peer-level review, what else can be done? Denham (Danaher) need comparator, utility
9:48am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Further training w/NA12878; pre-filtering var's. Still excludes indels, visual inspection still needed. #NGDx16
9:36am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Shows ROC 0.9978; 3FP 1FN out of 1664 datapoints, could reduce gene panel confirmations by 76%, but not good enough #NGDx16
9:35am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Shows complex desicsion-tree w/some 15 variables; QD, qual, Haplotype score, real depth, absRPRS etc #NGDx16
9:34am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: They have a conservative classifier to determine when Sanger isn't needed, via random forest on VCF file training #NGDx16
9:33am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: But to call indels correctly, QA measure, Sanger is the gold std. 'We are working hard to drop it however' #NGDx16
Bayrak-Toydemir: Sanger conf is hard to implement - doesn't scale, impacts TAT, not immune to its own problems #NGDx16
9:32am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: The NLGN4X vs NLGN4Y, only 2bp different; can't do it other than long-range PCR. #NGDx16
9:31am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Shows 68bp intronic insertion, with Sanger confirm. For Pseudoautosomal gene (NLGN4X), can't confirm via Sanger #NGDx16
9:30am August 23rd 2016 via Hootsuite
.@h2so4hurts Alas there is a 'red ocean' no matter what business you are in (i.e. for us it is Horizon Dx and homebrew...)
9:28am August 23rd 2016 via Hootsuite in reply to h2so4hurts
RT @michaelhoffman: @h2so4hurts @DaleYuzuki Never let pointlessness get in the way of a good tweet.
9:27am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Is the most damaging call (missense) the most likely one? #NGDx16
9:26am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Alignment and var calling errors - mid-size dels/dups: 10-20bp or even 50bp. Pseudogenes, pseudoautosomal #NGDx16
Bayrak-Toydemir: Failed run had 10x the unmapped read metric, causing suspicion, re-ran the samples #NGDx16
9:25am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Seq on flow-cell; can over- or underload; shows example of a HiSeq failure, all samples had same 2bp insertion #NGDx16
RT @h2so4hurts: Yeah, or you could just implement KING as a QC check. #NGDx16 https://t.co/5XTyxdxtbM https://t.co/O2f54YrhaP
9:24am August 23rd 2016 via Twitter Web Client
Bayrak-Toydemir: Looking at off-target reads, detected it. Only shown during separate capture events. Can't if it's the same capture #NGDx16
9:23am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Barcode contam: Looking at same region, two samples, same deletion (rare), cross-contam during opening strip-tubes #NGDx16
9:22am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Mentions use of GBlock (synth DNA) to help insure sample ID. Lib prep: human error contam.; PCR errors; Barcodes #NGDx16
9:20am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Mother and male fetus" heterozygous on the X-chr (sample mix-up). Difficult to fix, to understand how it happened #NGDx16
9:19am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Max errors is human error - mix-up, contam; extraction or quant also. #NGDx16
9:18am August 23rd 2016 via Hootsuite
Bayrak-Toydemir: Ti/Tv also. Focused on germline testing today. Sources of error: 'at every stage' - sample prep, lib, seq, bifx #NGDx16
Bayrak-Toydemir: Quality metrics - BAM; Variant metrics; Coverage. Focus - Q30 bases, not-callable, no-call regions, unmapped #NGDx16
9:17am August 23rd 2016 via Hootsuite
Pinar Bayrak-Toydemir (Med Dir, ARUP Laboratories) NGS in clinical diagnosis: aspects of quality management #NGDx16
9:14am August 23rd 2016 via Hootsuite
Q:How do you handle homopolymer regions? Zook: Descr but not able to assign confidence due to lack of good orthogonal methods #NGDx16 #GIAB
9:12am August 23rd 2016 via Hootsuite
Zook: #GiAB Registration is now open, Sept 15-16 Gaithersburg MD https://t.co/HTrWyD04bB #NGDx16
9:10am August 23rd 2016 via Hootsuite
Zook: Here is the #GIAB website https://t.co/AjavpepOo8 Next workshop Sept 15-16 at NIST (MD) #NGDx16
9:09am August 23rd 2016 via Hootsuite
Zook: Working on extending to SVs - callsets aren't sufficiently sensitive and imprecisely characterized #NGDx16 #GIAB
9:08am August 23rd 2016 via Hootsuite
Zook: Benchmarking: too many analytes (var's), calls biased towards easy var's and regions; manual inspection of FPs, FNs. #NGDx16 #GIAB
9:06am August 23rd 2016 via Hootsuite
Zook: Can engineer cell lines, synthetic DNA, simulate reads, modify real reads modify references. Or combination. #NGDx16 #GIAB
9:04am August 23rd 2016 via Hootsuite
Zook: Weakness of these reference materials - few deleterious var's, hard to sequence mutations. May collect add'l samples #NGDx16 #GIAB
9:03am August 23rd 2016 via Hootsuite
.@h2so4hurts No it isn't pointless! #NGDx16 needs your helpful and often-entertaining input! (NB - I'm not being sarcastic here.)
9:02am August 23rd 2016 via Twitter Web Client in reply to h2so4hurts
Zook: TP, FP, FN std def'n w/ @GA4GH and sophisticated benchmarking tools (vcfeval happ.py vgraph) https://t.co/HkPqJgZFXy #NGDx16 #GIAB
8:59am August 23rd 2016 via Hootsuite
Zook: Clinical app: preanalytical; fn annot; clin variant KB, clin interpretation; reporting; EHR achiving not addressed. #NGDx16 #GIAB
8:58am August 23rd 2016 via Hootsuite
Zook: WGS as a regulated clinical app? FDA as a novel medical device - two issues, broad, indication-blind testing; vast data #NGDx16 #GIAB
8:56am August 23rd 2016 via Hootsuite
Zook: The Sci Reports June '16 https://t.co/ANS1pFrHQy 51 authors; integrate methods. '14 ref https://t.co/XpvbUgOi8e #NGDx16 #GIAB
8:54am August 23rd 2016 via Hootsuite
Zook: NIST is 'bringing the principles of metrology to the genome'. Dev benchmarking tools with @GA4GH #NGDx16 #GIAB
8:48am August 23rd 2016 via Hootsuite
Zook: Lists multiple purposes: QC, QA, PT; source of bias; resolve hard SVs; improve ref assy's; enable regulated apps #NGDx16 #GIAB
8:46am August 23rd 2016 via Hootsuite
Zook: Materials for library prep through seq, alignment, var calling, confidence estimates. Not upstream or var interpret #NGDx16 #GIAB
8:45am August 23rd 2016 via Hootsuite
Zook: All 6 genome samples are consented for redistribution; mentions product can be derived (like @SeraCare using GM24385) #NGDx16 #GIAB
8:44am August 23rd 2016 via Hootsuite
Zook: Pilot genome NA12878; Ashkenazim Trio, Asian Trio (PGP) to be released Sept '16 Ref: https://t.co/kl1kAZ0gZf #NGDx16 #GIAB
8:42am August 23rd 2016 via Hootsuite
Zook: Alt title: "So you sequenced my genome, how well did you do?" Only 60% concordance Venn diag. https://t.co/tyuv3dWn74 #NGDx16 #GIAB
8:41am August 23rd 2016 via Hootsuite
Justin Zook (NIST) "Best practices for using Genome in a Bottle Reference Materials to Benchmark Variant Calls" #NGDx16 #GIAB
8:39am August 23rd 2016 via Hootsuite
Lohman: Answer: we need 'tons more information' - structural info, functional info. Much work is ahead to bring NGS to the clinic #NGDx16
8:35am August 23rd 2016 via Hootsuite
Kent Lohman from MRI Global (where @bioinformer works) starts the Clinical Assays track: 'why do we need more sequencing?' #NGDx16
8:34am August 23rd 2016 via Hootsuite
Okay, now I learn that the hashtag for this event is @NGDx16 https://t.co/bBw7RZX2Ty
8:18am August 23rd 2016 via Hootsuite
