Schilsky: 7 layers of security. PHI (privacy): '14 ASCO J ref https://t.co/9IGtogGG8Q #TRICON
2:46pm March 10th 2016 via Hootsuite
Schilsky: They are working on an NLP engine from unstructured data; but also a master pt index. Layered security. #TRICON
2:43pm March 10th 2016 via Hootsuite
Schilsky: Enables speed - for useful data at point-of-care. Can provide analysis within one second. (Amazing) #TRICON
2:42pm March 10th 2016 via Hootsuite
Schilsky: Data analysis overlaid; they are working with SAP's HANA Enterprise Cloud (HEC), in-memory db for rapid analytics #TRICON
Schilsky: Shows data flow into CancerLinQ, a common data-model in one part (so pt-level quality analytics) then de-identified #TRICON
2:41pm March 10th 2016 via Hootsuite
Schilsky: CancerLinQ brings in full EMR: nothing is pre-specified. Collected, mapped, eventually de-identified for many purposes #TRICON
2:39pm March 10th 2016 via Hootsuite
Schilsky: System is starting to bring in EMR from 6 practices; different than most registries (w/focused data collection) #TRICON
Schilsky: Built on 3 principles: non-exclusive technology, data governance stds (and privacy), and controlled data acess #TRICON
2:38pm March 10th 2016 via Hootsuite
Schilsky: Provide insights, hypothesis generation, ID the 'best evidence at the point of care'. Address interoperability in EMR too #TRICON
2:37pm March 10th 2016 via Hootsuite
Schilsky: ASCO's response: CancerLinQ - incorporating QOPI (quality measured) to build it into the system https://t.co/pzaQmBq4fL #TRICON
2:36pm March 10th 2016 via Hootsuite
Schilsky: Refers to this 2013 IOM report, 'Delivering high-quality care for a system in crisis' https://t.co/kYtTIisyTx #TRICON
2:35pm March 10th 2016 via Hootsuite
Schilsky: But compared to the general population, clinical trial participants are younger, more healthy, and less diverse #TRICON
2:34pm March 10th 2016 via Hootsuite
Schilsky: Decreasing numbers of particiapnts in clinical trials - due to better medications/treatments #TRICON
2:33pm March 10th 2016 via Hootsuite
Richard Schilsky (ASCO) Building a rapid learning system to improve cancer care #TRICON
Leary: BELLE-2 study, breast cancer study, PI3K/mTOR pathway, met primary endpoint for PFS (2 mos) n=1,147 #TRICON
1:04pm March 10th 2016 via Hootsuite
Leary: Chart of LoD, and regulatory compliance: qPCR and multiplex qPCR at the top, and capture / NGS at the bottom #TRICON
1:01pm March 10th 2016 via Hootsuite
Leary: Many types of questions - from research/discovery; exploratory/retrospective biomarker dev; to pt selection and clin trials #TRICON
1:00pm March 10th 2016 via Hootsuite
Rebecca Leary (Novartis) Assay characterization for clinical applications #TRICON
12:58pm March 10th 2016 via Hootsuite
Patel: At day 100 post-trtmtn showed spike in mutation in ctDNA, and PET scan positive. #TRICON
12:47pm March 10th 2016 via Hootsuite
Patel: Last case, subclinical recurrence of pancreatic cancer - 55d post-treatment (chemo/radiation), 15 copies/mL neg by PET #TRICON
Patel: Another case: TKI resistance via T790M after 1y of erlotinib trtmt. Liver lesion coincided with EGFR T790M, Ex19del in plasma #TRICON
12:45pm March 10th 2016 via Hootsuite
Patel:Case ex 1 Stg I Lung adenocarcinoma; high-dose radiation; KRAS G13D spikes; 50d - 150d undetect, 1y disease-free #TRICON
12:42pm March 10th 2016 via Hootsuite
Patel: A 'Sink' molecule controls for the entropy component. #TRICON
12:40pm March 10th 2016 via Hootsuite
Patel: Entropy vs Enthalpy (base-paring) - mutation shifts equilibrium, can raise the temp. But multiplexed? 'Sink' molecule... #TRICON
12:39pm March 10th 2016 via Hootsuite
Patel: Working with Dave Zhang (Searna) using Toehold probes '15 Nature Methods https://t.co/Td5DPJQ0qx to deplete WT sequences #TRICON
12:37pm March 10th 2016 via Hootsuite
Patel: cost of assay $20/1mL sample; 1 HiSeq lane, 100 samples. #TRICON
12:34pm March 10th 2016 via Hootsuite
Patel: Suppress PCR errors via molecular barcode - a 'lineage tag'. Can drive error down to <1:100,000. 50 hotspot regions #TRICON
12:33pm March 10th 2016 via Hootsuite
Patel: Able to get error down to 0.2% this way. '12 Cancer Res https://t.co/aQDvhsMZev #TRICON
12:32pm March 10th 2016 via Hootsuite
Patel: Via amplicon seq, going very deeply: address errors from the sequencer, and from PCR. Shows bi-directional sequencing #TRICON
12:30pm March 10th 2016 via Hootsuite
Patel: Mutation enrichment include COLD-PCR, LNA/PNA blocking, ARMS-PCR. Compartments via dPCR, BEAMing. (Contrasts to NGS) #TRICON
12:28pm March 10th 2016 via Hootsuite
Patel: Nevi contain BRAF mut's; large-scale studies may help determine what background ctDNA signals may be from healthy individs #TRICON
12:25pm March 10th 2016 via Hootsuite
Patel: Another problem w/protein biomarkers - can mask tumor signal, vary per-individual. ctDNA: specific, virtually no bkgnd #TRICON
12:24pm March 10th 2016 via Hootsuite
Patel: Shows graph of ref's w/'cell-free DNA' or 'circulating DNA' now at 350 in '15. Protein biomarkers: all have background #TRICON
12:23pm March 10th 2016 via Hootsuite
Abhijit Patel (Yale) Clinical applications of an NGS assay for ultrasensitive measurement of ctDNA #TRICON
12:20pm March 10th 2016 via Hootsuite
Q: Can you miss other mutations from original sample? Deihn: Yes but want to catch many of them; sampling many sites not feasible #TRICON
12:18pm March 10th 2016 via Hootsuite
Q: Detail around source of errors? Deihn:Oxidative damage is library-prep related; others relate to sequencer (will try alternative) #TRICON
12:14pm March 10th 2016 via Hootsuite
Deihn: Conclude 'high quality clinical studies are needed to prove clinical utility' #TRICON
12:13pm March 10th 2016 via Hootsuite
Deihn: Showed data of plasma across time; Dx mutations disappeared; example of pushing the limits further. #TRICON
12:12pm March 10th 2016 via Hootsuite
Deihn: 16yo GBM pt; trtmt surger, XRT, temosolomide, EGFRvIII vaccine; relapse day 485; 8 SNVs; cp Dx to relapse tumors #TRICON
12:11pm March 10th 2016 via Hootsuite
Deihn: Up the clinically relevant input: 750 genome equiv, 1500 mutations, predicted 0.00025%; found 2.5 molecules/1,000,000 #TRICON
12:09pm March 10th 2016 via Hootsuite
Deihn:'Like dPCR on steroids': pt-specific CAPP-Seq. Tumor biopsy WES/WGS; then develop a custom selector for ctDNA #TRICON
12:08pm March 10th 2016 via Hootsuite
Deihn:Built many CAPP-Seq selectors for many tumor types; validation sets from n=56 to 360. SKCM, PANC, DLBCL, FL, OV, BRCA etc #TRICON
Deihn: Add'l data of pre-treatment of cfDNA and mutations in tumor (right on top of each other) for NSCLC #TRICON
12:07pm March 10th 2016 via Hootsuite
Deihn: Benefit of a blood-based testing - being able to see the entire view. Showed high PPV of 94% #TRICON
12:05pm March 10th 2016 via Hootsuite
Deihn: Now applied to NSCLC: AUC is 1.0 for EGFR; all genes AUC is 0.94; believe it is due to heterogeneity of tumor biopsy #TRICON
12:04pm March 10th 2016 via Hootsuite
Deihn: iDES compared to other barcoding methods - too many to list, Schmitt et al 2012 and 2015 - it's the combination #TRICON
Deihn: Barcoding addresses stochastic; polishing address library errors; shows data that cfDNA recovery is optimized #TRICON
12:01pm March 10th 2016 via Hootsuite
Deihn: #Applying polishing or polishing - can drop error rate from 0.03% to 0.01%. iDES combines: 0.0015% #TRICON
12:00pm March 10th 2016 via Hootsuite
Deihn: But n=12;'15 Nature Comm https://t.co/8JdCdleVAE n=17 data supports model. Barcoding - ID's errors in lib prep #TRICON
11:59am March 10th 2016 via Hootsuite
Deihn: THus .05% may be practical lower limit without add'l work. Plot of all 16 error types (A>C, A>G etc) and freq; G>T common #T
11:57am March 10th 2016 via Hootsuite
