Bready: Nabsys maps can correct assembly errors, and do de novo, working on 'push-button de novo' combining NGS short-reads #AGBT14
2:15pm February 14th 2014 via Hootsuite
Bready: Illustrating a E. coli phylogenetic tree, and similarity between different strains (BL21, K-12 etc.) maps are dissimilar #AGBT14
2:13pm February 14th 2014 via Hootsuite
Bready: Workflow includes DNA extraction, probe attach, incubate with NPS reagent, load, analyze. #AGBT14
2:11pm February 14th 2014 via Hootsuite
Bready: Doing de novo assembly, simply lining up the tags. Instrument has 8 modules, independent of each other #AGBT14
2:10pm February 14th 2014 via Hootsuite
Bready: Showing deletion, translocation, inversion and how the maps would clearly & easily distinguish them #AGBT14
2:09pm February 14th 2014 via Hootsuite
Bready: Pressure- and voltage-based control; molecules of all sizes. 150ms traces - 150K bp, ~26 tags detected across the 150kb #AGBT14
Bready: Looks like a huge funnel at top, pass thru a single point, and then expand outward after the detection point #AGBT14
2:08pm February 14th 2014 via Hootsuite
Bready: Electronic data collection at high resolution, and long lengths. Electron micrograph of the detector #AGBT14
2:07pm February 14th 2014 via Hootsuite
Bready: Their readlength - ~100KB - sequence-spec tags, driven through a single detector at high velocity, 1MB/sec #AGBT14
2:06pm February 14th 2014 via Hootsuite
Bready: Long-range mapping tech: electronic has no diffraction limit like light has, much higher resolution #AGBT14
Bready: Illustrates their mapping ability as knowing the box picture for a puzzle; the 'big picture' that includes structural info #AGBT14
2:03pm February 14th 2014 via Hootsuite
First up: Barrett Bready (Nabsys) Correcting Genome Assemblies Using Semiconductor-Based, Single Molecule Genomic Mapping #AGBT14
2:02pm February 14th 2014 via Hootsuite
Myers: Q: Dazzler release? A: Only when it's ready, extract tool soon, but correction needs some more work (to make it easier). 1mo #AGBT14
12:08pm February 14th 2014 via Hootsuite
Myers: Q: Will you look at transcriptome assemblies? A: A hybrid strategy would be a good thing. May not have enough PacBio copies #AGBT14
12:07pm February 14th 2014 via Hootsuite
Myers: LR enables de novo, reference quality reconstructions, great for comparative genomics, and large scale variation #AGBT14
12:06pm February 14th 2014 via Hootsuite
Myers: Concl: high error 'is not a quality issue' - only efficiency and consensus. Eff problem solved here. #AGBT14
Myers: 2TB to 150GB filesizes using .dexta with PacBio human data #AGBT14
12:05pm February 14th 2014 via Hootsuite
Myers: G.bax.h5 'is a moose' of a file; fasta to .dexta down to 1/14th the size. #AGBT14
12:04pm February 14th 2014 via Hootsuite
Myers: Can correction be bypassed? A hard question. All pure strategies - PacBio only, then assemble. #AGBT14
12:03pm February 14th 2014 via Hootsuite
Myers:50x human, 512 cores, 5 days to assemble in 16GB. Uses DFS, will take 100x longer if not. #AGBT14
12:02pm February 14th 2014 via Hootsuite
Myers: Only FASTAs needed for input, quality comes in after for consensus for Quiver. Shows E.coli, Arabidopsis, Human assy times #AGBT14
Myers: 36x faster - Dazzle vs. Blaser, more sens, catches same true positives. Overlap, scrub, correct, overlap, scrub, assemble #AGBT14
12:00pm February 14th 2014 via Hootsuite
Myers: You don't know when the PacBio error spikes (technology problem) but after consensus correction, to 0.5% error rate at 10x #AGBT14
11:59am February 14th 2014 via Hootsuite
Myers: Don't think about the graph; scrub the underlying data. "This is for the gear-heads". (Comment: great talk nonetheless) #AGBT14
11:57am February 14th 2014 via Hootsuite
Myers: Illustrates a string graph problem - contiguous stretches of underlying genome (untrue with real-world data), and error rate #AGBT14
11:56am February 14th 2014 via Hootsuite
Myers: String graphs are the real problem. His paper from 2005 http://t.co/h4tlst8Fn0 #AGBT14
11:54am February 14th 2014 via Hootsuite
Myers: 20x of 15% data gives Q70; the problem is the ability to assemble efficiently. Punch-line: 'We're 37x faster" on a modest h/w #AGBT14
11:53am February 14th 2014 via Hootsuite
Myers: At perfect Poisson, would get 100% coverage; 'this is why I'm excited' with PacBio data #AGBT14
11:52am February 14th 2014 via Hootsuite
Myers: Implications - the higher x, the accuracy will converge geometrically with depth. #AGBT14
Myers: Talking about RS 2 - P4/C2, 11% insertion, 3% deletion, 1% subst. But errors are random; read sampling also >nearly< random #AG
11:51am February 14th 2014 via Hootsuite
Myers: After a 10y hiatus, knew short-reads were not 'intellectually satisfying to me' (in the meantime) #AGBT14
11:50am February 14th 2014 via Hootsuite
Gene Myers, Max Planck Institute for Molecular Cell Biology and Genetics: A de novo Whole Genome Shotgun Assembler for Noisy LR Data #AGBT
11:49am February 14th 2014 via Hootsuite
McCombie: Q: What about >7.5kb cutoff? A: Not that many really long reads; need new isolation procedures to get unbroken DNAs #AGBT14
11:48am February 14th 2014 via Hootsuite
McCombie: Q: (Mardis) At 100x, is it purifying selection? A: 100x total coverage, >10kb reads a lot lower at 20x or so #AGBT14
11:47am February 14th 2014 via Hootsuite
McCombie: Arabidopsis strain Ler-0 sequenced, also O. sativa.10kb mean RL, max 54kb (!) #AGBT14
11:44am February 14th 2014 via Hootsuite
McCombie: Also did S. pombe dg21, max RL 35.4kb, mean 5.17kb, 275x in 5 SMRTcells "huge difference in loading efficiency" for PacBio #AGBT14
11:39am February 14th 2014 via Hootsuite
McCombie: Rice genomes about 450MB, wanted to try using new PacBio chemistry (P5-C3), uses yeast W303 strain; only >10kb fx used #AGBT14
11:37am February 14th 2014 via Hootsuite
Up next: W.R. McCombie, Cold Spring Harbor Laboratory: A Near Perfect de novo Assembly of a Eukaryotic Genome Using >10KB Reads #AGBT14
11:34am February 14th 2014 via Hootsuite
Jaffe: Q: Cp to PacBio? A: "Hard-pressed to find a method to get to zero errors". Something to be said for size, cost, portability #AGBT14
11:33am February 14th 2014 via Hootsuite
Jaffe: Q: K-mer disambiguation had two poly-G's A: Doesn't make sense for long homopol's 'from the data I've seen' #AGBT14
11:32am February 14th 2014 via Hootsuite
Jaffe: Q: What if it doesn't get better? A: Not quite the right Q; a large % of the bases are right, do sth else in the lab. #AGBT14
11:31am February 14th 2014 via Hootsuite
Jaffe: Q:(Mardis) Misleading title? A: Assembly would have introduced "a bunch of errors", so we did something that was useful. #AGBT14
11:30am February 14th 2014 via Hootsuite
Jaffe: Path forward is to improve base calling, and to mix pore types. #AGBT14
11:29am February 14th 2014 via Hootsuite
Jaffe: Small device, 5kb median, 10kb tail, long perfect stretches common; 'error clumps' repeatedly at one loci #AGBT14
Jaffe: Resolving ambiguities with ONT data one long repeat and tandem repeat. #AGBT14
11:28am February 14th 2014 via Hootsuite
RT @neilhall_uk: OK so far we can see that this data is not going to replace PACBIO in the near future. #AGBT14
11:26am February 14th 2014 via Hootsuite
Jaffe: Alternate paths, score them, can resolve the graph. 13x for Scardovia, smaller 1+ MB genome. Showed short-read assy #AGBT14
11:25am February 14th 2014 via Hootsuite
Jaffe: Just combining datatypes - use 250b ILMN data, use DISCOVAR for assembly. Explains edge and node, illustrates reads laid over #AGBT14
11:24am February 14th 2014 via Hootsuite
Jaffe: Some consistent artifacts - missing T-base (5/6 E. coli reads with missing base.) Could improve the algorithm; mix pore-types #AGBT14
11:22am February 14th 2014 via Hootsuite
Jaffe: Lots of good k-mers - how to think of this datatype. Showed segment of Scardovia reads #AGBT14
