Deignan: Have been using QC metrics from the sequencer as a control for the assay. Minor deviations can be handled #FDANGSWorkshop
12:16pm February 25th 2016 via Twitter Web Client
Q: Panel subset? Other changes? Klees: Has to make sure the assay will not be affected. #FDANGSWorkshop
12:14pm February 25th 2016 via Twitter Web Client
Klees: Has to validate the entire assay, so there are some cases where only a BAM would be sufficient. #FDANGSWorkshop
12:12pm February 25th 2016 via Twitter Web Client
Roscoe: Should validation controls include BAMs? Process controls? Van Allen: So many factors, just dry-bench not enough #FDANGSWorkshop
12:11pm February 25th 2016 via Twitter Web Client
Deignan: Tumor-only is what they've used; and the question of what is 'normal'. There are many germline challenges #FDANGSWorkshop
12:08pm February 25th 2016 via Twitter Web Client
Roscoe: Somatic vs Germline vars: reported separately? Drawbacks? Van Allen: Careful with the FPs, many variable to look at #FDANGSWorkshop
12:05pm February 25th 2016 via Twitter Web Client
Eberhard: Which quality metric dimensions are part of medicine, and which are assay specific. #FDANGSWorkshop
12:02pm February 25th 2016 via Twitter Web Client
.@VanAllenLab Glad my 'open notebook' could be of help!
12:00pm February 25th 2016 via Hootsuite in reply to VanAllenLab
Hedge: Agreed, 'definitely not Sanger', low VAF across sample types would be more useful than going 'outside' #FDANGSWorkshop
11:57am February 25th 2016 via Twitter Web Client
Roscoe: What is the appropriate orthogonal validation for accuracy? Deignan: "Not Sanger", perh another NGS, or another smpl #FDANGSWorkshop
11:55am February 25th 2016 via Twitter Web Client
Hegde: Focus on seq depth: but context of sequence complexity impt, also depth cutoffs #FDANGSWorkshop
11:51am February 25th 2016 via Twitter Web Client
Hegde: With indels a bioinformatic approach can be very useful. #FDANGSWorkshop
11:49am February 25th 2016 via Twitter Web Client
Hegde: May be able to design a synthetic probe. Consider how assay is designed, validate against true positives #FDANGSWorkshop
11:48am February 25th 2016 via Twitter Web Client
Q:Quantifying indel performance? Van Allen: Make sure you don't miss the well-char, actionable indels. Somatic BRCA2 indels #FDANGSWorkshop
11:46am February 25th 2016 via Twitter Web Client
Eberhard: Intended Use context - where do you want to go with the assay? On tumor cellularity: is there an internal metric? #FDANGSWorkshop
11:43am February 25th 2016 via Twitter Web Client
Deignan: But how many EGFR Exon 19 deletions exist? How many smpls is enough? #FDANGSWorkshop
11:40am February 25th 2016 via Twitter Web Client
Deignan: Clin relevance may not be that impt in this context; much intrinsic variants in any smpl. Ex: EGFR Exon 19 del #FDANGSWorkshop
11:39am February 25th 2016 via Twitter Web Client
Hegde: A critical and sens. Q for a mfr. Numbers? Adequate testing (sens, spec) but how difficult to get smpl types #FDANGSWorkshop
11:37am February 25th 2016 via Twitter Web Client
Roscoe: To eval performance, what variants? Challenging parameters? Platform bias? How to represent this? LoD? #FDANGSWorkshop
11:35am February 25th 2016 via Twitter Web Client
Donna Roscoe (FDA): Seeing input on analytical perf with a representative var set may infer for an entire panel #FDANGSWorkshop
11:34am February 25th 2016 via Twitter Web Client
Van Allen: Mentions the fact that intense interest in immuno-oncology and neoantigens, will be missed by panels #FDANGSWorkshop
11:32am February 25th 2016 via Twitter Web Client
Van Allen: Inferring global genome properties from panel testing: from 300 gene panel and inferring mutational load #FDANGSWorkshop
11:31am February 25th 2016 via Twitter Web Client
Van Allen: Analysed many germline db's (including ExAC) 157 clinical exomes, modeled 300 genes, got 14% FP rate #FDANGSWorkshop
11:27am February 25th 2016 via Twitter Web Client
Van Allen: #FDANGSWorkshop The '15 Science Trans Med ref: https://t.co/AmfNNWZuf3
11:26am February 25th 2016 via Twitter Web Client
Van Allen: Tumor-only analysis - germline FP mutations. Velculescu at Hopkins reported 30-50% FPs are germlines #FDANGSWorkshop
11:25am February 25th 2016 via Twitter Web Client
Van Allen: Same BAM, 3 different approaches, only 2 of 100's overlap #FDANGSWorkshop
11:24am February 25th 2016 via Twitter Web Client
Van Allen: Contrasting to fusion detection: RNA based, showed a messy Venn diagram of prostate samples diff analysis #FDANGSWorkshop
11:23am February 25th 2016 via Twitter Web Client
Van Allen: Somatic mutation detection vs fusion det: point mutation calls MuTect paper https://t.co/CKrrGvchjH #FDANGSWorkshop
11:22am February 25th 2016 via Twitter Web Client
Eli Van Allen (Dana Farber CC) NGS Panels and bioinformatics strategies for oncology #FDANGSWorkshop
11:20am February 25th 2016 via Twitter Web Client
Hegde: Sample pooling needs to be addressed in validation. "Not a huge problem' 0.18% in library prep, 0.03% subsequent #FDANGSWorkshop
Hegde: LOD, sens and spec are 'absolutely critical'. If it cannot detect <10% 'it must be disclosed'. #FDANGSWorkshop
11:18am February 25th 2016 via Twitter Web Client
Hegde: Reprod and versatility with different specimen types, reporting add'l needs. #FDANGSWorkshop
11:16am February 25th 2016 via Twitter Web Client
Hegde: About 20 smpls for analytical validation, ~1000x coverage; VAF freq <20%. Strategic mixtures to get VAF 10% or less #FDANGSWorksho
Hegde: Approaches to validation: involve hetergeneous tissues, evidence-based selection of targets, VAF. #FDANGSWorkshop
11:14am February 25th 2016 via Twitter Web Client
Hegde: At 10^11 bases/run, systematic mutation detection of all mutation types, clear advantages of NGS #FDANGSWorkshop
11:12am February 25th 2016 via Twitter Web Client
Hegde: And SNV detection and indel / other mutation types (CNVs, SVs) bioinf approaches #FDANGSWorkshop
11:11am February 25th 2016 via Twitter Web Client
Hegde: Many points to consider re:Targeted panels. Capture method, specimen diffs, detection methods/filters #FDANGSWorkshop
11:09am February 25th 2016 via Twitter Web Client
Hegde: Ability for HTP sequencing to look at SNV, CNAs (indels, translocations). Drug repurposing for new cancer types #FDANGSWorkshop
11:08am February 25th 2016 via Twitter Web Client
Hegde: Points out difference between cancer vs rare disease: use of RNA and DNA, re-sampling #FDANGSWorkshop
11:07am February 25th 2016 via Twitter Web Client
Also R Klees (NY State DOH) First up Madhuri Hedge (Emory) Germline testing for mosaicism similar to oncology #FDANGSWorkshop
11:06am February 25th 2016 via Twitter Web Client
Panel 2: Analytical Validation and Bioin #FDANGSWorkshop M Hegde (Emory) E Van Allen (Dana Farber) J Deignan (UCLA) D eberhard (UNC)
11:04am February 25th 2016 via Twitter Web Client
A (Aisner): Tried unk smpls that needed orthogonal validation. Will use a small number of unk smpls, to guard against bias #FDANGSWorkshop
10:37am February 25th 2016 via Twitter Web Client
Q2: Commutability - do clin samples have to have a var? How many? A: (Pfeifer) >100 pt smples. Revalidation: dozens #FDANGSWorkshop
10:35am February 25th 2016 via Twitter Web Client
Pfeifer (con't): Have to stop talking about formalin fixation as the 'evil'; focus may be too narrow. One of may variables #FDANGSWorkshop
10:33am February 25th 2016 via Twitter Web Client
Pfeifer: Want reprod and accurate testing; formalin fixation time standardization may need to be done. (Horses, barns) #FDANGSWorkshop
10:32am February 25th 2016 via Twitter Web Client
Aisner (con't) Pathology at that level is a completely different level - many constraints, not feasible #FDANGSWorkshop
10:31am February 25th 2016 via Twitter Web Client
Open floor:Q: Could nucl acid sample claims be met by education of pathologist? A:(Aisner) - diverse pathologists #FDANGSWorkshop
10:30am February 25th 2016 via Twitter Web Client
Berger: Like the idea of in-silico validation, Since they can't test every possible variant. Mixing doesn't do CNV #FDANGSWorkshop
10:28am February 25th 2016 via Twitter Web Client
Luthra: Does the cell line reflect real-life samples? Heme malignancies - still need cell lines. Need large # of pt smples #FDANGSWorkshop
10:27am February 25th 2016 via Twitter Web Client
Pfeifer: Mentions dry-lab tests to look at bioinformatic sensitivity in analysis #FDANGSWorkshop
10:26am February 25th 2016 via Twitter Web Client