BH: Across HLA-DPA1, HLA-DPB1, HLA-DPB2. Not their focus but can give a glimpse of potential #AGBT15
4:54pm February 27th 2015 via Hootsuite
BH: Can go in and pull out only the exome, at 139x, can get 95% SNPs phased; 96% of genes #AGBT15
4:53pm February 27th 2015 via Hootsuite
BH: (That video actually got applause. Nicely done!) #AGBT15
4:52pm February 27th 2015 via Hootsuite
BH: Built Loupe, a haplotype-aware browser. Animation from NA12878: nothing like a cool video with a waltz. 'Who's your daddy?' #AGBT15
4:51pm February 27th 2015 via Hootsuite
BH:Software based on std tools (BWA, GATK), linux-based, 'open-platform for high-throughput NGS' #AGBT15
4:49pm February 27th 2015 via Hootsuite
BH: Recycling - going back to the template with additional adapter priming/extension rounds. Then finish ends w/PCR #AGBT15
4:48pm February 27th 2015 via Hootsuite
BH: <10fg / partition; 0.3% of genome per GEM; 90% gel bead fill rate. #AGBT15
4:47pm February 27th 2015 via Hootsuite
BH:Gelbead dissolves, releasing thousands of barcodes into the bubble. So 1ng DNA now has >100K barcoded partitions #AGBT15
4:46pm February 27th 2015 via Hootsuite
BH: 54um bead, movie looks just like RainDance with emulsifying in oil. >100K reactions/5min. Gelbead in emulsion ("GEM") #AGBT15
4:45pm February 27th 2015 via Hootsuite
BH: GemCode system. 750K reagents in 1 tube; a 14bp barcode with a gelbead; highly uniform; built-in seq adapter & primer #AGBT15
4:44pm February 27th 2015 via Hootsuite
BH:Why not done at scale before? 384 partitions, Now >100K. Barcodes 384 vs. 750K. Input 100ng vs 1ng #AGBT15
4:43pm February 27th 2015 via Hootsuite
BH: Instead of nanopore, individual partitioning and molecular barcode. For both WGS and targeted; simple workflow #AGBT15
4:42pm February 27th 2015 via Hootsuite
BH: With linked reads: Accurate SNPs, haplotyping, precise counting, critical content (HLA), de novo, structural var's #AGBT15
4:41pm February 27th 2015 via Hootsuite
BH: 'This is the first technical presentation on 10X ever.' #AGBT15
4:40pm February 27th 2015 via Hootsuite
Ben Hindson 10X Genomics “Long Range Applications With Short Read Sequencing” #AGBT15
JS: Used BAC-based approach to get closure of Chr20 gap. #AGBT15
1:37pm February 27th 2015 via Hootsuite
JS: 'We discovered a large number of large struct var's in genes that are disease related' #AGBT15
1:33pm February 27th 2015 via Hootsuite
JS: For struct var, used Symap, UCSC tools, then a script to compare to GRCh38. Est total in AK1 ~21K 130kb ins 79kb del #AGBT15
JS: Used optical maps (Bionano) to fill gaps, also SMRT subreads. Of 604 total, 250 closed, 180 extended, 174 still open Total 5.4Mb #AGBT15
1:27pm February 27th 2015 via Hootsuite
JS: AK1, used 215GB ~72x on PacBio; ave RL 13.4kb. FALCON + Daligner made it fast - 4-5d. N50 7.3Mb, 2.8Gb length. 91% coverage #AGBT15
1:25pm February 27th 2015 via Hootsuite
JS: Mongolian project (GENDISCAN), Kazakshstan Genome project, Korean BAC Clone project (2001) #AGBT15
1:23pm February 27th 2015 via Hootsuite
Jeong-Sun Seo (Seoul Nat'l Univ, MacroGen) "De novo assembly of an asian diploid genome using SMRT sequencing" #AGBT15
1:21pm February 27th 2015 via Hootsuite
DC: Can do 1st and 2nd-pass alignments in GRCh37 and compare to GRCh38; beware of misassembly of 37 that appears as missing in 38 #AGBT15
1:15pm February 27th 2015 via Hootsuite
RT @drgitlin: DC: alternate loci contain genes! If you’re not using them in your analyses you’re missing important information. #AGBT15
1:12pm February 27th 2015 via Hootsuite
RT @AndyLarrea: Fear the Franken alleles #agbt15 @deannachurch http://t.co/G6k9KCvaQK
DC: 100kb deletion in this CCL3 region. Now onto GChr37, burst of activity in alternative loci and genes. Many seq's under-annotated #AGBT15
DC: CCL3 region - CNV could affect HIV susceptibility; but genotype is incorrect. CHM1 data ref: http://t.co/hiYW4BMzzA #AGBT15
1:09pm February 27th 2015 via Hootsuite
DC:The prior model was compressing 2 haplotypes into one that wasn't reflective of reality. New model: represent both w/alt locus #AGBT15
1:06pm February 27th 2015 via Hootsuite
DC: (And errors in reference confuse or dilute that variant signal - only gene 1 is 'present' in reference) #AGBT15
1:05pm February 27th 2015 via Hootsuite
DC:Need high-quality ref to call variants - and annotate them. Illus w/paralogous gene, where sample has deleterious var in gene 2 #AGBT15
DC: @personalisinc w/ACE WES can solve 50% of cases (good) but what about the other 50%? Var ID but not matched to phenotype #AGBT15
1:03pm February 27th 2015 via Hootsuite
Deanna Church (Personalis) "Finishing genomes: why does it matter?" #AGBT15
1:01pm February 27th 2015 via Hootsuite
GM: Briefly goes through idea of scrubbing (cleaning up artifacts in reads); module DAscrub. Illus beautifully w/ 30x Ecoli data #AGBT15
12:59pm February 27th 2015 via Hootsuite
GM: "I pulled every trick I ever learned" with Daligner to get 25x - 40x speed improvement over BLASR #AGBT15
12:57pm February 27th 2015 via Hootsuite
GM: http://t.co/DRwyr36HqT is where his code lives. Lossless compression at 14x, can analyze compressed; Daligner module fast #AGBT15
12:56pm February 27th 2015 via Hootsuite
GM: An assembler is a pipeline of modules; HGAP, Falcon; plug-and-play. It takes time to build good modules for all components #AGBT15
12:55pm February 27th 2015 via Hootsuite
GM: Current assemblers 'already competitive'. Melanogaster using PB 100x or Ref 6 (hand curated) comparable #AGBT15
12:54pm February 27th 2015 via Hootsuite
GM: Surprised w/lack of assembler. Not restricted by error rate; need was efficient in seq coverage & computer time w/15% error #AGBT15
12:53pm February 27th 2015 via Hootsuite
GM: Sufficiently var repeats are resolvable. As long as you can span. 1 het per every 2kbp for P6/C4 #AGBT15
12:52pm February 27th 2015 via Hootsuite
GM:Thus - just a function of coverage 3) Spanned repeats are resolvable. Unique flank on both sides solves it http://t.co/kX2c8cxvh7 #AGBT15
12:50pm February 27th 2015 via Hootsuite
GM: 2) Read error is Random - Q score (consensus of k sequences) increases linearly in k #AGBT15
12:47pm February 27th 2015 via Hootsuite
GM: 1)Poisson: At any desired level of coverage, amt of coverage approaches 100% (Lander-Waterman 1988) #AGBT15
GM:As a mathematician - near-perfect assy back on the table. After 10y hiatus - "I'm back" @thegenemyers #AGBT15
12:46pm February 27th 2015 via Hootsuite
GM: 10kbp+ length (P6/C4), 10-15% error (mostly insertions), but: random 'at any given point'. Read sampling also random. #AGBT15
12:45pm February 27th 2015 via Hootsuite
Gene Myers: #AGBT15 My favorite talk from 2014 here: http://t.co/YY0rHw7cGO
12:44pm February 27th 2015 via Hootsuite
JV:Q:What abt epigenetics? A:We need the equiv for the methylome - HTP and accurately. But: epigenetics still encoded by genome #AGBT15
12:42pm February 27th 2015 via Hootsuite
JV:Insurance co's are not interested in prevention, long-term health; large companies though that self-insure, do. #AGBT15
12:41pm February 27th 2015 via Hootsuite
JV: 'I know 5 technologies down at the single molecule level' 'We're counting on $30 genomes in 3-4y' (seriously) #AGBT15
12:40pm February 27th 2015 via Hootsuite
RT @chrisamiller: Finishing up the paper and dbGaP submissions now! MT @DaleYuzuki Hope to make it avail 'soon' #AGBT15
12:38pm February 27th 2015 via Hootsuite
JV:Goal is 1M genomes + phenomes by 2020. Genomics moving forward at a 'much more rapid pace in a much more predictive fashion' #AGBT15
12:37pm February 27th 2015 via Hootsuite
