Chris Wright Profile
Chris Wright

@chrisnrg

Followers
491
Following
289
Statuses
716

Oxfordshire
Joined October 2009
Don't wanna be here? Send us removal request.
@chrisnrg
Chris Wright
5 months
@mattloose worst bug report eva
1
0
2
@chrisnrg
Chris Wright
5 months
@mattloose reports that his iPad isn't a fan of running the kraken2 classifiers, so I'd be interested to hear about other people's experience and get some bug reports.
1
0
1
@chrisnrg
Chris Wright
5 months
@krannich479 Medaka never used TF directly, only through keras (back when keras supported multiple frameworks the first time). That was because TF was a massive PITA to use on its own --- and kinda still is. Which is where we land at torch.
0
0
3
@chrisnrg
Chris Wright
10 months
@BonfieldJames Intriguing! I say this having written a noddy index for BAM to allow subsequent fetching of blocks of a fixed number of reads. It's a common thing we find ourselves doing to allow parallelism in workflows.
0
0
0
@chrisnrg
Chris Wright
11 months
@samstudio8 @nanopore Version 2.0, bold move.
0
0
2
@chrisnrg
Chris Wright
1 year
@bioinfomatt @NeilHorner @nanopore It's almost like you were trying to get featured in some sort of weekly newsletter.
1
0
2
@chrisnrg
Chris Wright
1 year
@BonfieldJames I wrote some dumb programs with kseq.h to front @nanopore EPI2ME workflows to similar effect. It's robust to a variety of weird inputs and so can regularise data for all the downstream tools. (Somewhat against my better judgement to go fix other people's tools)
0
0
0
@chrisnrg
Chris Wright
1 year
@bioinfomatt I should never have put you two in charge of anything.
0
0
2
@chrisnrg
Chris Wright
1 year
@RobAboukhalil @bioinfomatt @nanopore We have some unloved notebook tutorials here: I don't fancy compiling some of the tools used to wasm though!
1
1
1
@chrisnrg
Chris Wright
1 year
@chrisnrg
Chris Wright
1 year
@kirk3gaard @nanopore We've written about how to perform such corrections previously If squiggly maths scares you: all this is saying is "add on the average error rate to all reads"
Tweet media one
0
0
3
@chrisnrg
Chris Wright
1 year
@Psy_Fer_ @gringene_bio Labelling reads with a 0.5 is simply a non-informative prior. If you don't like the formulae I posted previously for how to use the data to finesse that idea, look up additive smoothing the result is much the same.
1
0
1
@chrisnrg
Chris Wright
1 year
@jpelbers @Psy_Fer_ @msikic @JiazhiLiu391571 @nanopore James, I can only think this is bait for me to post maths formulas again....
2
0
4
@chrisnrg
Chris Wright
1 year
RT @bioinfomatt: We’re expanding the team with new junior analyst roles. Help us develop new and exciting Nanopore end to end workflows for…
0
5
0
@chrisnrg
Chris Wright
1 year
@kirk3gaard @nanopore Not quite so bluntly, see the last equation. If you set alpha=0.5 and w=1, then that's "adding on" an error of half-a-base to all reads. You can play with these parameters (cue outrage from some); the equations tell you how to set them using the data itself.
0
0
4
@chrisnrg
Chris Wright
1 year
@Psy_Fer_ @kirk3gaard @nanopore @pypi Sure, if it were proven to be more powerful than medaka. We stopped development of diploid calling in medaka when @aquaskyline was pushing harder than we could.
0
0
0