Slater: ML predict, act and adapt. Deep learning is a subset, training and using nural networks as a predictive model #genomicsfest
1:31pm October 4th 2017 via Hootsuite
Slater: Analytics vs Machine Learning (ML): statistics create a model of the world. Take data, run model, reiterate in ML #genomicsfest
1:30pm October 4th 2017 via Hootsuite
Slater: Defined: AI has been around for a while; 'just using a computer to simulate an intellectual task' #genomicsfest
1:28pm October 4th 2017 via Hootsuite
Ted Slater (Cray) The what, how and why of deep learning in the life sciences #genomicsfest Asks for show of hands who is using DL
1:27pm October 4th 2017 via Hootsuite
LaFranzo: Still in development but looks like a promising approach; look at TMB from RNA-seq data from FFPE. #genomicsfest
12:27pm October 4th 2017 via Hootsuite
LaFranzo: Tumor Mut Burden (TMB) for neoantigenicity; matched/normal; now via RNA data, can select genes and provide estimate #genomicsfest
12:26pm October 4th 2017 via Hootsuite
LaFranzo: Shows mixtures of RNAs and degradation experimental data. Also IHC comparison as Hi/Med/Lo quant of markers #genomicsfest
12:25pm October 4th 2017 via Hootsuite
LaFranzo: Lab has databases of immune cell profiling; an RNA fingerprint to deconvolve the mixture of cells in TME #genomicsfest
12:22pm October 4th 2017 via Hootsuite
LaFranzo: For tumor microenv (TME): define Pt populations for targeted Rx; cand ID new prognostic, diagnostic markers #genomicsfest
12:19pm October 4th 2017 via Hootsuite
LaFranzo: Went through CAP for RNA; Pinnacle was their first offering for FFPE and RNA. Limited material; unpredictable. #genomicsfest
12:18pm October 4th 2017 via Hootsuite
LaFranzo: CoFactor is both biotechnology and bioinformatics. Started as an NGS service provider; focus is on RNA in '13 #genomicsfest
12:17pm October 4th 2017 via Hootsuite
Natalie LaFranzo (Cofactor Genomics) Diving deep into the tumor microenvironment #genomicsfest
12:16pm October 4th 2017 via Hootsuite
Patel: Ab with oligos - with UMI's on antibodies, simultaneous RNA and protein. #genomicsfest
12:14pm October 4th 2017 via Hootsuite
Patel: Panels of 400 targets: breast cancer, T-cell expression, immune response, stem cells. In dev: immuno-oncology #genomicsfest
12:10pm October 4th 2017 via Hootsuite
Patel: Claim higher sens, better resolution, economy of sequencing (60% reads are dominated by high-abundance xcrpts) #genomicsfest
12:09pm October 4th 2017 via Hootsuite
Patel: Magnets pull beads in bulk, synthesize cDNA, sequence. Targeted approach: panels of pre-selected genes #genomicsfest
12:06pm October 4th 2017 via Hootsuite
Patel: Magnetic beads are 1:1, constrained by Poisson. Well has single cell and single bead; lyse cells and captured on bead #genomicsfest
12:05pm October 4th 2017 via Hootsuite
Patel: 200K microwells, sparse loading. Easy-to-use (compared to emulsion droplets). See 65% load to capture efficiency #genomicsfest
Patel: 'Bringing single-cell RNA-seq to any lab'. Powering high number of samples; 20K cells via gravity deposition. 50um #genomicsfest
12:04pm October 4th 2017 via Hootsuite
Patel: A microwell capture-based system. Beads to capture RNA, targeted gene expression for better sens, decrease in costs #genomicsfest
12:02pm October 4th 2017 via Hootsuite
Patel: Acquisition of Cellular Research for single-cell RNA analysis, molecular indexing of RNAs. Rhapsody is targeted GEx #genomicsfest
12:01pm October 4th 2017 via Hootsuite
Kieren Patel (BD_ Rhapsody and AbSeq - RNA and protein analysis #genomicsfest
12:00pm October 4th 2017 via Hootsuite
Q: future of NovaSeq? Schweitzer: Almost too big for what they do. RNA-Seq; too high a multiplex doesn't make a lot of sense #genomicsfest
11:36am October 4th 2017 via Hootsuite
Schweitzer: Other key advantage is turnaound-time. They still have big Sanger operation; same-day service. #genomicsfest
11:33am October 4th 2017 via Hootsuite
Q: How to respond to outsourcing? Schweitzer: At times they do refer outside. Value of a core: someone to get consultation #genomicsfest
Schweitzer: Secondary indexing primers, 384 samples, grew out of VitusGen project. (No details on the multiplex tech) #genomicsfest
11:28am October 4th 2017 via Hootsuite
Schweitzer: Sequencing costs approaching zero? Driving down cost of library prep for large projects - AmpSeq, 500-plex #genomicsfest
11:27am October 4th 2017 via Hootsuite
Schweitzer: Shows size-selection using Sage BluePippin and AATi characterization in prep for 10X Genomics libraries #genomicsfest
11:26am October 4th 2017 via Hootsuite
Schweitzer: Their use of AATii fragment analyzer and Bio-Rad ddPCR used extensively for library QC. Shows data from femtopulse #genomicsfest
11:24am October 4th 2017 via Hootsuite
Schweitzer: Criteria for new instrumentation is 'workhorse': particularly useful, and with multiple purposes. #genomicsfest
11:16am October 4th 2017 via Hootsuite
Schweitzer: 'Genomics for the rest of us': VitisGen2 to improve wine and table grapes (USDA), Gates Found. RepGen for crops #genomicsfest
11:14am October 4th 2017 via Hootsuite
Schweitzer: Do a lot of non-human work. Birds, plants, veterinary medicine in Ithaca (horses and sharks), even entomology #genomicsfest
11:12am October 4th 2017 via Hootsuite
Peter Schweitzer (Cornell Univ) Implementing new technologies for genomes and cutting-edge research in a shared facility #genomicsfest
11:10am October 4th 2017 via Hootsuite
Kellis: W/ full clinical record: a huge spectrum. What will systematic mining show? What underlying biology will be revealed? #genomicsfest
10:36am October 4th 2017 via Hootsuite
Kellis: Phenotype is 1000's of datapoints per individual. GWAS has trad been done with a single phenotype. #genomicsfest
Q: Ethical standpoint? Pickrell: A fundamental ethical one is 'should the pt get their own data?' #genomicsfest
10:35am October 4th 2017 via Hootsuite
Zimmer: Academic usage? Gerstein: Students dig into Carl's genome as an individual; powerful. 'Real people' impt connection #genomicsfest
10:27am October 4th 2017 via Hootsuite
Gerstein: A question is how this will be monetized; could be via insurance. #genomicsfest
10:26am October 4th 2017 via Hootsuite
Sutti: But GINA doesn't protect against long-term insurance in the US. Will be a problem in the future #genomicsfest
Q: Genes and role in court system, insurance? Imptc of counselors? Sutti: Walk through risks/benefits WRT test. GINA reviewed #genomicsfest
10:25am October 4th 2017 via Hootsuite
Kellis: A huge amt to learn - env and genetics and phenotypic info; epigenetics can give that bridge to contextualize #genomicsfest
10:24am October 4th 2017 via Hootsuite
Kellis: As diet changes, genetic effects may show. Thus dynamic; info can be gathered, env and phenotypic data collected #genomicsfest
10:23am October 4th 2017 via Hootsuite
Pickrell: Have multiple groups giving their best effort, and then compete for the best answer. Sutti: need for std testing #genomicsfest
10:20am October 4th 2017 via Hootsuite
Q: How to communicate info? Need to standardize? Pickrell: "no" Gerstein: At least try to std-ize, large-scale statistics #genomicsfest
10:19am October 4th 2017 via Hootsuite
Pickrell: Spend time asking about family history, understand what people think they are asking, to get the connection #genomicsfest
10:18am October 4th 2017 via Hootsuite
Kellis: Potential in banking info, samples, over time (and starting a company around this idea). Genetic hints, biol knowledge #genomicsfest
10:16am October 4th 2017 via Hootsuite
Kellis: Bridging the gap between what can be modulated; epigenetic changes; systematic measurements possible. #genomicsfest
10:15am October 4th 2017 via Hootsuite
Kellis: "Not everyone will need to go through what you Carl went through' Points out the danger of determinism. #genomicsfest
10:14am October 4th 2017 via Hootsuite
Zimmer: Points out somatic mosaicism is real. A huge response, as people have image of genome as 'sacred, unchanging object' #genomicsfest
10:12am October 4th 2017 via Hootsuite
Neat Game of Genomes panel discussion at #genomicsfest with @joe_pickrell @manoliskellis @markgerstein and hosted b… https://t.co/Yrn9FPoTGe
