James Bonfield
banner
jbonfield.bsky.social
James Bonfield
@jbonfield.bsky.social
Walker, archer, and volunteer woodland warden by weekend, and bioinformatics software engineer and general geek by weekday.

My favourite prime is 15551, my favourite colour is, obviously, octarine, and I love nothing more than being immersed in nature.
A feeling of dirtiness, a sense of horror, verging on revulsion. So maybe it's a perfect fit for VCF? 🤣

My gut feeling is we'd be better just having a JSON equivalent to VCF than JSON inside VCF. (Eg if we wanted JSON, then the whole of info should be json, not just each element.)
November 7, 2025 at 8:29 AM
I hear you wish -S. It's not so much that it's not needed, it doesn't harm, but that it's typically indicative of out of date information still doing the rounds/

My pet peeve is "samtools view -b foo.sam > foo.bam" instead of "-o". It causes lots of confusion with nohup and similar.
October 30, 2025 at 1:44 PM
I've heard that before too, but the naming of it predates my involvement so I've no idea if it stood for anything. It's plausible, but *shrug*.
October 30, 2025 at 8:12 AM
You can't out reform Reform. They're loony conspiracy theory nutters - you shouldn't *want* to woo them!

We need parties that accept why the economy has tanked and will take steps to fix it. Leaving the ECM - Common Market - was the dumbiest idea ever (regardless of how you voted on Brexit).
October 9, 2025 at 4:47 PM
I noted in their presentation they said that samtools mpileup didn't work. I think they're a bit out of date. Bcftools mpileup --poly-mqual can handle the qualities in homopolymers, plus other newer -X profiles.

I haven't tuned it yet though for SBX, but think it'll be OK in general. (To try!)
September 18, 2025 at 8:47 PM
Instagram is like facebork but even more annoying. I looked and I can't even find the equivalent post for you over there. It's just a hateful platform. Probably OK for doom scrolling on a phone, but that's about it.
September 16, 2025 at 8:01 AM
We ought to update htslib.org with more precise recipes, especially for things like conda where we know A) people make mistakes, often and B) it's used A LOT. We may be able to point to something like biocontainers too (or roll our own, but I'd rather not).

It's rarely built from source it seems.
Samtools
Samtools
htslib.org
September 15, 2025 at 1:58 PM
Even more prolific is looking at their WhatsApp number from the minimap2 fake site, and associated email. So so many fake sites. Scary

(See 447950904740 phone number, and emmawatsofficial54 partial email search results).
September 15, 2025 at 9:06 AM
A google for the support phone number shows how many other phishing sites they have.

www.google.com/search?clien...

Most likely their "support" offering involves getting you to install some trojan.
Google Search
www.google.com
September 15, 2025 at 8:53 AM
Nothing like cold hard data. It's almost as if Brexit was a pack of lies? Who'd have believed it. ;-)

Of course the people that need to see this obviously won't as it'll be deemed "fake news". I really don't know how to fix that one.
September 14, 2025 at 1:19 PM
The binary version changes are probably the biggest issue, with (IIRC) BCF 4.2 not being readable by bcftools and BCF 4.3 not being readable by GATK, as the minor version bump was a breaking change that made them incompatible.

I think it was necessary as some data was broken, but :-( :-( :(
August 8, 2025 at 8:37 PM
FWIW if I ever get time to finish my bgzf2 (zstd) branch (github.com/jkbonfield/h...) it really shines with multi-sample VCF.

The line lengths are just too big for bgzf to do remotely well due to the 32Kb deflate window size.
GitHub - jkbonfield/htslib at bgzf2
C library for high-throughput sequencing data formats - GitHub - jkbonfield/htslib at bgzf2
github.com
August 8, 2025 at 8:29 PM
BCF gets some things right, but it made many of the same mistakes that BAM did (being of the same era). It's too serial rather than block based, harming any sort of efficient processing and compression. In short, it's the binarisation of the text format that makes it poor.
August 8, 2025 at 8:28 PM
A text format we can hack and play with allows for fast experimentation, but it shouldn't be the primary format. Not should we have binary guys which are essentially memory dumps from parsing the text. That's partially what killed BCF from adoption. All the pain with minimal gain!
August 7, 2025 at 6:29 AM
I've not tried this before. Thanks

🧩 Puzzle #735

🤔 22 guesses

⏱️ 6m 43s

🔗 alphaguess.com
alphaguess: a word game
Guess the word of the day. Guesses reveal where the word is positioned alphabetically. Everyone plays the same word each day.
alphaguess.com
July 29, 2025 at 7:40 PM