August 2016

Stephanie Shulman (NY State) "The biggest issue is with intended use, with little or no data". They've been reviewing LDTs for years #NGDx16

3:46pm August 25th 2016 via Hootsuite

Putcha: Looking at screening or MRD, the NIPT analogy 'only goes so far' as they have not gone thru FDA. 'The bar is very high' #NGDx16

3:02pm August 25th 2016 via Hootsuite

Casdin: Competition - illustrates Sequenom as a first-mover, once the fastest-growing (70% penetration y1) but didn't make it #NGDx16

2:53pm August 25th 2016 via Hootsuite

Eli Casdin (Casdin Capital) says that the most valuable portion of the diagnostic business is their distribution (sales) channel #NGDx16

2:52pm August 25th 2016 via Hootsuite

de Abreu: Data storage: cloud has higher latency, but local is large investment, maint., and limited disaster recovery options #NGDx16

11:49am August 25th 2016 via Hootsuite

de Abreu: For completeness of coverage, Ion CHPv2 expect 2 drop-outs; TruSeq dependency of DNA quality (<3 or >3; higher=worse) #NGDx1

11:47am August 25th 2016 via Hootsuite

de Abreu: Some companies will do end-to-end they've tried: PierianDx, QCI Analyze/interpret, N of One (partial) #NGDx16

11:42am August 25th 2016 via Hootsuite

de Abreu: 3 QC data analysis steps; Increasing manual process post-variant calling. Filtering, manual seq rev; curation v impt #NGDx16

11:40am August 25th 2016 via Hootsuite

de Abreu: Workflow on pre-post extraction; library prep and data analysis are starred, where QC steps occur #NGDx16

11:36am August 25th 2016 via Hootsuite

de Abreu: CGAT is Dartmouth's new name (Clinical Genomics and Advanced Technology). Illustrates move from transl to clinical #NGDx16

11:35am August 25th 2016 via Hootsuite

Francine Blumental de Abreu: Advantages and challenges of detecting molecular biomarkers in oncology: data analysis perspective #NGDx16

11:32am August 25th 2016 via Hootsuite

Mockus: 82K variants, in publicly-available database here: https://t.co/KDevr0EM58 #NGDx16

11:30am August 25th 2016 via Hootsuite

Mockus: Nice '16 paper on clinical trial landscape and the JAX-CKB (clinical KB) https://t.co/gPp0J3BhIM #NGDx16

11:24am August 25th 2016 via Hootsuite

Mockus: These issues raised in recent Clin Chem rev https://t.co/015RDVCIB9 #NGDx16

11:22am August 25th 2016 via Hootsuite

Mockus: What about del exon 19, or p.E746_A750del? And could be no clinical data, preclinical level or full clinical #NGDx16

11:20am August 25th 2016 via Hootsuite

Mockus: Illustrates the 'specificity spectrum' where the mutation is poorly described ("EGFR mutation"); esp in registries. #NGDx16

11:19am August 25th 2016 via Hootsuite

Mockus: Their KB is in the middle - curated, admits a data silo, challenge in how to best share this data #NGDx16

11:14am August 25th 2016 via Hootsuite

Mockus: DB levels: CGVR, GMDR, GMED (clinical genomic var repository; genomic medical data repos; genomic medical evidence db) #NGDx16

11:13am August 25th 2016 via Hootsuite

Mockus: Their KB then combines matched disease and variants: gene var, Rx w/efficicacy and ranking, clinical trials results, status #NGDx16

11:09am August 25th 2016 via Hootsuite

Mockus: Their eLIMS not only uses MeSH, but also Disease ontology from UMd from Path rpt to match disease. Var caller for match vars #NGDx16

11:08am August 25th 2016 via Hootsuite

Mockus: Not only reproducibility, but traceability to pull data back out again. #NGDx16

11:06am August 25th 2016 via Hootsuite

Mockus: Shows diagram clin analytics -> actionable vars -> Rx and clinical trials -> clinical decisions. Accuracy, speed, reprod #N

11:04am August 25th 2016 via Hootsuite

Mockus: For a comprehensive db: published somatic var's; relevant Rx w/supporting efficacy evidence; and clinical trials #NGDx16

11:02am August 25th 2016 via Hootsuite

Mockus: Why use large panels? WES? WGS? A debate for a while until costs come down and technology for coverage improves #NGDx16

11:01am August 25th 2016 via Hootsuite

Susan Mockus (Jackson): Challenges and solutions with interpretation of large genomic panels. ~11M genomic publicaitons; 200/day #NGDx16

10:57am August 25th 2016 via Hootsuite

Daber: Overlapping PCR primers via separation is trad method; now will change primer placement to optimize for med/larg indels #NGDx16

9:36am August 25th 2016 via Hootsuite

Daber: Can the wet lab capture the event; can the informatics pick it up? Limitations of amplicons w/dropout; while hyb capture can #NGDx16

9:34am August 25th 2016 via Hootsuite

Daber: Last topic on indels; out of unique var's, 1/4 (1117/5279) are indels. AML, MPN, larger mutations #NGDx16

9:33am August 25th 2016 via Hootsuite

Daber: 3rd is to use anchored multiplex (ArcherDx) but needs 1kb-2kb region w/unique sequence. But need a priori identification #NGDx16

9:32am August 25th 2016 via Hootsuite

Daber: Second strategy is to remove homologous region; 100% specific reflex assay, to get greater specificity #NGDx16

9:31am August 25th 2016 via Hootsuite

Daber: Using enzymatic shearing, had a problem with insert size being too small and losing anchor. #NGDx16

9:30am August 25th 2016 via Hootsuite

Daber: Regions of homology: using anchoring regions with PE reads, but depends if library insert size is consistent #NGDx16

9:30am August 25th 2016 via Hootsuite

Daber: From 12K samples, 2% of reported var have been <5% AF. 'It is critical to broaden validation claims in lower allele burdens' #NGDx

9:28am August 25th 2016 via Hootsuite

Daber: 200ng vs 1ng vs 0.5ng - look at VAF sens and spec data across mutations and mutation types #NGDx16

9:27am August 25th 2016 via Hootsuite

Daber: They can drill down, allele freq titration at different DNA inputs; look at reprod at 0.5%, 1%, 3% AF. #NGDx16

9:26am August 25th 2016 via Hootsuite

Daber: Then look at true events, can detect SNV 0.5% AF, indels 2-5bp 4%, indels 6-15bp 1%, indels >15bp 0.2% and NPM1 insert 0.5%. #NGDx

9:25am August 25th 2016 via Hootsuite

Daber: Looking at JAK2 technical performance >12K samples; Looking at an intrinsic rate of eror <0.3% for A base; <0.35% overall #N

9:23am August 25th 2016 via Hootsuite

Daber: Pushing LOD - '12 down to 5%; in '16, tools are not flexible to define LOD at each spec genomic location #NGDx16

9:22am August 25th 2016 via Hootsuite

RT @h2so4hurts: TS: Axiom array identifies over 12K different microbes. Rapid platform. Great tool for the right application #NGDx16

9:21am August 25th 2016 via Hootsuite

Daber: Need to create artificial FASTQ files '13 ref https://t.co/Lgd3US39YK Shows algorithms to optimize/validate pipeline #NGDx16

9:21am August 25th 2016 via Hootsuite

Daber: Informatics and library prep are connected; some assays will have better sens than NGS; 2-20bp indels (med), 20-200bp (large) #NGDx16

9:19am August 25th 2016 via Hootsuite

Daber: IHC still impt - PD-1/L1. Technologies involve mutation, FISH, IHC; samples are small in number, multiple library preps #NGDx16

9:17am August 25th 2016 via Hootsuite

Daber: Personalized med - ALK+ crizotinib, EGFR+ afatinib, otherwise std Rx. Also chart of NCCN biomarkers by cancer type #NGDx16

9:15am August 25th 2016 via Hootsuite

Daber: Last topic - intermediate and large <200bp indels, esp with amplicon-based capture #NGDx16

9:14am August 25th 2016 via Hootsuite

Daber: Can we personalize var det across the entire genome? Some regions more complex than others; strategy for homologous regions #NGDx16

9:13am August 25th 2016 via Hootsuite

Bob Daber (VP Genomic R&D BioReference) Pushing the limits: challenges of somatic variant detection #NGDx16

9:12am August 25th 2016 via Hootsuite

Wetterskog: Leverage heterozygous SNPs, to ID mono-allelic deletions (LOH) in low tumor content (10%). #NGDx16

8:49am August 25th 2016 via Hootsuite

Wetterskog: Study n=97, 274 samples, Royal Marsden and IRST (Italy). Targeted NGS, 39kb, common dels and SNVs. #NGDx16

8:47am August 25th 2016 via Hootsuite