Stegle: Applied to 45K retina cells via DropSeq: reactome 387 processes, top one is 'hidden' explaining variability #ESHG2016
3:31am May 22nd 2016 via Hootsuite
Stegle: On second point - 'modeling zero inflation and dropout' - zero data is just not enough read depth; get imputed expression #ESHG2016
3:30am May 22nd 2016 via Hootsuite
Stegle: On challenge is computational cost of number of cells (from 100's to 100K). Second is cell counts vs seq depth #ESHG2016
3:29am May 22nd 2016 via Hootsuite
Stegle: This factorial single-cell latent var model '10 ref https://t.co/fGUN8qYpUL Gene sets allow discovery of best-fit marker #ESHG2016
3:26am May 22nd 2016 via Hootsuite
Stegle: Cell differentiation, other confounding factors: generalizing to decompose the gene matrix (cells x genes) #ESHG2016
3:24am May 22nd 2016 via Hootsuite
Stegle: 27% of genes, var can be explained by technical noise. Shows correction of the cell cycle effect. #ESHG2016
3:23am May 22nd 2016 via Hootsuite
Stegle: Naive to Th2, Th1 etc: cell cycle has larger-than-expected impact on transcriptional variation than expected #ESHG2016
3:22am May 22nd 2016 via Hootsuite
Stegle: The covariance matrix is at single-cell level; correcting for factors that do not matter (apoptosis, cell cycle). #ESHG2016
3:21am May 22nd 2016 via Hootsuite
Stegle: '15 Nat Genetics Rev https://t.co/NRKi06U7XT Gene expression heterogeneity is not new; fit via covariance matrices #ESHG2016
3:20am May 22nd 2016 via Hootsuite
Stegle: Observed exp profiles 'do not enable recovering of differentiation process' #ESHG2016
3:18am May 22nd 2016 via Hootsuite
Stegle: Many add'l biological sources of variation (cell cycle, apoptosis...) Example: hidden variation look at Gata3 exp in Th2 #ESHG2016
Stegle: Single-cell RNA-seq is now 'routine'. Look at novel variation between cells; can get transcriptome-wide var #ESHG2016
3:17am May 22nd 2016 via Hootsuite
Stegle: In the context of quantitative genetics, regulatory genomics, and single-cell heterogeneity. #ESHG2016
3:16am May 22nd 2016 via Hootsuite
Oliver Stegle (EBI Wellcome Trust UK) Modeling genetic and non-genetic sources of variation in single cells #ESHG2016 @oliverstegle
3:15am May 22nd 2016 via Hootsuite
Q: Timing of expression? Gilad: Some ideas to perform at the single-cell level, possible in the future #ESHG2016
3:09am May 22nd 2016 via Hootsuite
Gilad: Recommend: multiplex samples to avoid confounding batch effect to avoid bias #ESHG2016
3:06am May 22nd 2016 via Hootsuite
Gilad: Shows PCA, correcting for batch effect, that it matters a lot. (corrected clusters much better) #ESHG2016
3:05am May 22nd 2016 via Hootsuite
Gilad: Even when ERCCs added in bulk - but 'you have to think about it for a while' why ERCCs are affected by native RNA #ESHG2016
3:04am May 22nd 2016 via Hootsuite
Gilad: Shows 8 batches; use ERCCs to correct for this? Didn't work, as ERCCs affected even more batch-to-batch (!) #ESHG2016
3:03am May 22nd 2016 via Hootsuite
Gilad: There are differences in batches; conversion of molecules vs reads for each cell shown in a color graph; data shown #ESHG2016
3:02am May 22nd 2016 via Hootsuite
Gilad: Mean gene exp in bulk vs single: at 70-80 cells, recapitulate 90% of genes in bulk. #ESHG2016
3:01am May 22nd 2016 via Hootsuite
Gilad: ERCC and mean molecule count behaves 'slightly differently' than endogenous genes. Justifies use of UMI's #ESHG2016
3:00am May 22nd 2016 via Hootsuite
Gilad: (UMI - unique molecular identifier barcode). They are looking at getting around batch effects on C1. #ESHG2016
2:59am May 22nd 2016 via Hootsuite
Gilad: 'Choice between Fluidigm C1 vs DropSeq; decided to use C1, 'eventually everyone may use DropSeq'. Describes use of UMI's #ESHG2016
2:57am May 22nd 2016 via Hootsuite
Gilad: Changing gears completely, now on single-cell data from all samples. Not LCLs but iPSCs #ESHG2016
2:56am May 22nd 2016 via Hootsuite
Gilad: #EID splicing QTLs, mapping junction reads between introns/exons LeafCutter tool pre-print: https://t.co/aF5MYs2Lrz #ESHG2016
2:55am May 22nd 2016 via Hootsuite
Gilad: eQTLs - across the regulatory cascade, >60% of regul. variants impact gene regulation through chromatin enhancers #ESHG2016
2:51am May 22nd 2016 via Hootsuite
Gilad: Impact of regulatory variation - more at RNA, ribosomal level; effect drops off at protein level called 'buffering' #ESHG2016
2:48am May 22nd 2016 via Hootsuite
Gilad: Work involves how to visualize 5 variables plus H3K27ac (distal enhancers). Now at TSS (promoter), shows downstream effect #ESHG2016
2:46am May 22nd 2016 via Hootsuite
Gilad:Shows a sample heatmap of H3K27ac (TSS), 4sU (30m, 60m), RNA-seq ("G" and "P"), Ribo, protein: pair-wise comparison check #ESHG2016
2:45am May 22nd 2016 via Hootsuite
Gilad: The majority of the time: data analysis, batch effects, age, sex, date, tech, sequencer, sample quality etc etc. #ESHG2016
2:43am May 22nd 2016 via Hootsuite
Gilad: N.B. - here's a '14 rev https://t.co/9mbFwgO9PS on eQTLs and regulatory mechanisms #ESHG2016
2:42am May 22nd 2016 via Hootsuite
Gilad: 'Regulatory cascade': H3K27ac ChIP, DNA methyl, DNAse-seq, 4sU-seq (transcription rate), RNA-seq, ribo-seq, MassSpec #ESHG2016
2:40am May 22nd 2016 via Hootsuite
Gilad: Studying 70 human, 8 chimps, 8 macaques: relates many dimensions across 8y: gene exp, chromatin accessibility, etc etc #ESHG2016
2:38am May 22nd 2016 via Hootsuite
Gilad: Distal epigenetics, promoter epigenetics, xcr rate, stale mRNA, translation, stable protein levels in complex disease #ESHG2016
2:37am May 22nd 2016 via Hootsuite
Gilad: Mapping of steady-state, mature levels of mRNA. But what about other components of transcription machinery? #ESHG2016
2:36am May 22nd 2016 via Hootsuite
Gilad: Shows expression level variation of different individuals with different genotypes: allele-spec, so typically cis-effects #ESHG2016
2:35am May 22nd 2016 via Hootsuite
Gilad: Many GWAS hits are in gene deserts - regulatory regions. Thus eQTL link variation to changes in gene regulation #ESHG2016
2:34am May 22nd 2016 via Hootsuite
Oops, never mind about McCarroll. Yoav Gilad Univ Chicago "RNA splicing is a primary link between genetic variation and disease" #ESHG2016
2:33am May 22nd 2016 via Hootsuite
Good morning #ESHG2016 - starting with 'Understanding functional effects of genomic variants' with McCarroll and Stegle single-cell RNA-Seq
2:31am May 22nd 2016 via Hootsuite
RT @TCGAupdates: #NCICCG Blog -- Insights & Innovations in #cancergenomics https://t.co/4p9QPcYRsf https://t.co/JkJoHckYYB
12:53am May 22nd 2016 via Hootsuite
RT @TCGAupdates: Molecular Subtyping of #OvarianCancer wit Multi-omics Data #openaccess https://t.co/pYciWBvDM2 https://t.co/aZWACdiDUx
10:45pm May 21st 2016 via Hootsuite
Protective gene offers hope for next blockbuster heart drug : Nature News https://t.co/W6Ej7LB6oO
9:50pm May 21st 2016 via Hootsuite
Chinese officials 'create 488m bogus social media posts a year' | The Guardian https://t.co/aNGEO5GsHQ
8:40pm May 21st 2016 via Hootsuite
This six-minute short film plunges you into an augmented reality hellscape | The Verge https://t.co/rYTRrf8Flu
7:00pm May 21st 2016 via Hootsuite
HIV cure a step closer after scientists remove virus's DNA from living tissue | The Independent https://t.co/g2W0TWRtWs
6:40pm May 21st 2016 via Hootsuite
Peek Inside Tri Alpha Energy, a Company Pursuing the Ideal Power Source | Technology Rev https://t.co/PZpUBMhzQU
5:40pm May 21st 2016 via Hootsuite
CDC monitoring nearly 300 pregnant women with Zika in U.S. states, territories - The Washington Post https://t.co/HjiQFgyQar
4:40pm May 21st 2016 via Hootsuite
Exp. validation of methods for differential gene expression analysis and sample pooling in RNA-seq | BMC Genomics https://t.co/o52LqntXlg
3:40pm May 21st 2016 via Hootsuite
Lopez-Otin: Describes his work with CLL in conjunction with ICGC; ID'd the relevance of NOTCH1 in CLL #ESHG2016
9:36am May 21st 2016 via Hootsuite
