Q: Det cancer type? What genes, types of muts? Reddi: Looked at 2y of gene-panel results; checked against literature #genomicsfest
3:42pm October 4th 2017 via Hootsuite
Crompton: Li-Fraumeni, is 60% risk, could use liquid biopsy for freq monitoring. Problem to know what to look for, potential #genomicsfest
3:40pm October 4th 2017 via Hootsuite
Q: Liquid biopsies used proactively (screening)? Crompton: No data at all - 'very excited about it'. Pts with predisposition #genomicsfest
Reddi: QIAGEN putting resources into preanalytical variables. What is the best method as 'truth'. #genomicsfest
3:39pm October 4th 2017 via Hootsuite
Chevillet: Raises a good point - where skin cells are accidently collected in a blood draw, where epithelial CTC markers used #genomicsfest
3:38pm October 4th 2017 via Hootsuite
Thorner: Standardization of collection? Reddi: Starting-point can all be arguable; tumor cellularity in FFPE? #genomicsfest
3:36pm October 4th 2017 via Hootsuite
Crompton: Can imagine the day where resistance is actively monitored, managed like HIV. #genomicsfest
3:35pm October 4th 2017 via Hootsuite
Q: Where is the most interesting dev? Crompton: For ctDNA, the clinical prognostic outcome. #genomicsfest
3:34pm October 4th 2017 via Hootsuite
Chevillet: Unk where the most useful biomarker will be - in exosomes, in CTCs, in ctDNA. 'We have to show it.' #genomicsfest
3:32pm October 4th 2017 via Hootsuite
Chevillet: But CTC selection is a problem. MicroRNAs thought to be encapsulated and then it wasn't true. #genomicsfest
Crompton: How to control for CTCs in ctDNA samples? At 0.1%, it all matters. Chevillet: Ways to select for / against CTCs #genomicsfest
3:30pm October 4th 2017 via Hootsuite
Reddi: Not easy to transfer NIPT over to cancer; many more considerations. #genomicsfest
3:29pm October 4th 2017 via Hootsuite
Thorner: Could it be used upfront? Crompton: Need to know clinical goal. ctDNA in all tumor types except for transloc (>50%) #genomicsfes
3:26pm October 4th 2017 via Hootsuite
Crompton: Ped ca driven by transloc and CNV; not hotspots. Thornton: How looking at CNV? A:WGS and Broad CNV caller #genomicsfest
3:25pm October 4th 2017 via Hootsuite
Chevillet: Tec-Seq paper to exceed error rate - barcode molecules, amplify, '17 ref https://t.co/AVXsLZtHZD #genomicsfest
3:23pm October 4th 2017 via Hootsuite
Reddi: Also plasma samples from collaborators. Thorner: How to exceed limit of accuracy of NGS? Reddi: use ddPCR to confirm #genomicsfest
3:21pm October 4th 2017 via Hootsuite
Reddi: They have to consider patient needs, foremost. Thorner: how is LOD est? A: Known samples, from SeraCare or Horizon. #genomicsfest
3:20pm October 4th 2017 via Hootsuite
Reddi: Their lowest input was 1mL of plasma, 10ng of cfDNA. At 20,000x coverage, set LOD at 1% from a clinical utility persp #genomicsfest
Chevillet: Lots of different problems - monitoring therapy, screening. May well have multiple methods of interrogation. #genomicsfest
3:18pm October 4th 2017 via Hootsuite
Thorner: How do you know you have enough cfDNA? Crompton: his patients 'are small'. Pairing blood draw w/apheresis, not real #genomicsfest
3:16pm October 4th 2017 via Hootsuite
Brian Crompton (Dana Farber) Pediatric oncologist, runs a translational research lab. Not engineers, implementing est tech #genomicsfest
3:13pm October 4th 2017 via Hootsuite
John Chevillet (Draper Labs MA) Spun out of MIT years ago, dev microfluidic tech for CAR-T therapy, currently $500K #genomicsfest
3:11pm October 4th 2017 via Hootsuite
Reddi: two cancer panels, a fusion panel, near final validation of a liquid biopsy assay. Partnerships with academics, pharma #genomicsfest
3:10pm October 4th 2017 via Hootsuite
Panel: Honey Reddi (Jackson Labs) John Chevillet (Draper) Brian Crompton (Dana Farber) Reddi: Jackson 'one campus w/o mice' #genomicsfest
3:09pm October 4th 2017 via Hootsuite
Thorner: His group optimizes assays, and deploys them into the clinic. Sequenom 10y ago; Oncopanel for NGS #genomicsfest
3:08pm October 4th 2017 via Hootsuite
Aaron Thorner (Cancer Genome Discovery, Dana Farber) Realizing the potential of liquid biopsies: are we ready? #genomicsfest
3:07pm October 4th 2017 via Hootsuite
Q: Cofactor had 14 investors at seed. Ideal number? Messina: Quality over quantity; those truly invested in you #genomicsfest
2:16pm October 4th 2017 via Hootsuite
Q: How did you find investors to talk with? Messina: people you know; companies list who invest in them; also investor conf's #genomicsfest
2:13pm October 4th 2017 via Hootsuite
Messina: Do you really need slides? You want engagement. It's a 2-way conversation. You are looking to enter a relationship. #genomicsfest
2:10pm October 4th 2017 via Hootsuite
Messina: 'If you are doing most of the talking, you are doing it wrong.' You need to hear from them as much as you say. #genomicsfest
2:09pm October 4th 2017 via Hootsuite
Messina: Talk about the team most; the market next; then a little about the product 'in any real detail'. #genomicsfest
2:08pm October 4th 2017 via Hootsuite
Messina: "Pitch your grandma" - be able to talk about what you are doing in plain and simple language #genomicsfest
2:06pm October 4th 2017 via Hootsuite
Messina: Y-combinator only gives you 10 minutes. If they don't 'get it at the beginning', well you waste a lot of time #genomicsfest
Messina: Talking to investors: get to the point 'right away'; put all the good stuff up-front. #genomicsfest
2:05pm October 4th 2017 via Hootsuite
Messina: Crowdfunding is non-dilutive, but difficult to do. Cofactor bootstrapped by friends and family at the beginning #genomicsfest
2:03pm October 4th 2017 via Hootsuite
Messina: VCs and institutional investors are $500K and up; don't be afraid to ask. Book rec: https://t.co/cdE3KPhB3d #genomicsfest
1:59pm October 4th 2017 via Hootsuite
Messina: Sources: friends and family. Limited amts, need to sign a contract; Angel investors; VCs #genomicsfest
1:57pm October 4th 2017 via Hootsuite
Messina: Series A, $20M led by Y-combinator and Stanford. Will talk about different funding sources, how to talk to investors #genomicsfest
1:54pm October 4th 2017 via Hootsuite
Messina: One of the first CLIA labs in the country for RNA assays. Paragon assay is their immune resp in tumor microenv #genomicsfest
1:53pm October 4th 2017 via Hootsuite
Dave Messina (Cofactor Genomics): How to fund your biotech startup #genomicsfest
1:52pm October 4th 2017 via Hootsuite
Slater: Expect ROI payoff in 1-2y; 'time to begin experimentation is now'. (This from an HPC provider, of course...) #genomicsfest
1:48pm October 4th 2017 via Hootsuite
Slater: Training time to best accuracy is currently 5d; in 2-5y, 2h. Hardware is 25GFlops now; expect TFlops. #genomicsfest
1:47pm October 4th 2017 via Hootsuite
Slater: (More information about the CANDLE project here:) https://t.co/oNN3DL8yWx #genomicsfest
1:46pm October 4th 2017 via Hootsuite
Slater: Large project - CANDLE for exascle deep learning - Ras pathway, prediction of drug response, Argonne Nat'l Lab #genomicsfest
1:44pm October 4th 2017 via Hootsuite
Slater: "HPC and AI will converge" - Gartner quote - 'AI and ML have reached a critical tipping point...' #genomicsfest
1:42pm October 4th 2017 via Hootsuite
Slater: Behind the scenes, it's about model training and model testing. More data run through models #genomicsfest
1:41pm October 4th 2017 via Hootsuite
Slater: Shows workflow chart of DL in prod: Training/Inferencing/Model Mgmt. Data preparation is a key bottleneck. #genomicsfest
1:36pm October 4th 2017 via Hootsuite
Slater: NVIDIA person quote "Deep learning in healthcare is the leading industrial application of AI" #genomicsfest
1:35pm October 4th 2017 via Hootsuite
Slater: Other trends - compute power, adv algorithms, data science expertise. In life sciences: diagnosis, pt stratification #genomicsfest
1:34pm October 4th 2017 via Hootsuite
Slater: Shows a History of DL slide; electronic brain in 1940; Perceptron in 1957; ADALINE '60. Why DL now? So much data avail #genomicsfest
1:32pm October 4th 2017 via Hootsuite
