Sign in

Bede Constantinides

@bede.im
1K followers 1.6K following 213 posts

Research fellow at the University of Birmingham interested in infectious disease informatics. Also cycling, photography, active travel. bede.im

PostsRepliesMedia
Bede Constantinides @bede.im · 29/09/2026
"An estimated 2.3 million new cancer cases were attributable to infections globally in 2024… Equivalent to 12% of all cancer cases" www.thelancet.com/journals/lan...
thelancet.com
Global burden of cancer attributable to infections in 2024: a worldwide incidence analysis
Our findings highlight the importance of infection control to achieve cancer prevention. Scientifically proven, but often underused, approaches include HPV and HBV vaccination; testing and treatment o...
031
Bede Constantinides @bede.im · 27/09/2026
Celebrated road cycling world champs with a mates race of ~20 riders in a double breakaway scenario. Podium secured for team Norway inc @nickdsandy.bsky.social 🇳🇴
152
Bede Constantinides @bede.im · 25/09/2026
Big AI's black box harnesses seem great in practice for research, presumably helped by having fast access to a cache of all published research. But ofc they move under one's feet and are tricky to benchmark.
000
Bede Constantinides @bede.im · 25/09/2026
Yes, OpenAI models refuse bio prompts less often than Anthropic. That's quick! Presumably via chatgpt.com in Instant mode or similar rather than a harness? gpt-6-sol high takes 4mins through this setup, I'll add it shortly.
100
Bede Constantinides @bede.im · 24/09/2026
"Claude Opus 5.5 refused to answer on biosecurity grounds, its refusal costing half that of DeepSeek’s complete answer"
061
Bede Constantinides @bede.im · 24/09/2026
Another reason I'm excited about this launch is that @pathoplexus.org is using Deacon to classify human sequences in submissions to help avoid inadvertent disclosure of human genetic information github.com/bede/deacon
000
Bede Constantinides @bede.im · 24/09/2026
Josh (@scalene.bsky.social) and I asked the best open-weight models a tricky molecular biology research question that Opus 5.5 refuses. Josh graded each answer, and DeepSeek V4 Pro and Kimi K3 both beat Opus 5, taking less time and costing less. Blogged: bede.im/2026/09/24/o...
bede.im
Asking open models about molecular biology
293
Reposted by Bede Constantinides
Conor (Conchúr) Meehan @conmeehan.bsky.social · 23/09/2026
We are looking for someone with a strong background in metagenomics and epidemiology to join our team for ~2 years at UKHSA to work on the mSCAPE programme. Please see (and share) the link below and feel free to message me with any enquiries. www.healthjobsuk.com/job/Canary_W...
healthjobsuk.com
Job vacancy: Senior Scientist (Epidemiology), UK Health Security Agency, Canary Wharf, London | trac.jobs
0510
Bede Constantinides @bede.im · 23/09/2026
…The status quo in virus genomics is to share only assembled genomes. While precious, these consensus genomes bake in biases of lab and informatic processing, leading to systematic errors in virus phylogeny: www.nature.com/articles/s41... With raw reads, many of these errors can be fixed!
nature.com
Addressing pandemic-wide systematic errors in the SARS-CoV-2 phylogeny - Nature Methods
This Resource paper presents a global SARS-CoV-2 phylogenetic tree of 4,471,579 high-quality genomes consistently constructed by Viridian, an efficient amplicon-aware assembler.
111
Bede Constantinides @bede.im · 23/09/2026
This is a big deal for virus data sharing. Submitters can now upload raw reads, immediately share them with collaborators, and let Pathoplexus take care of the INSDC submission process on their behalf…
1101
Reposted by Bede Constantinides
Pathoplexus @pathoplexus.org · 22/09/2026
1/ 🧬 A big #Pathoplexus milestone: you can now optionally include raw sequencing reads alongside your consensus sequences! Hosting raw reads alongside consensus data helps improve reproducibility & allow in-depth analyses. 📰 Read more: pathoplexus.org/news/2026-09...
AI-generated image showing raw reads (short reads) lining up above a consensus sequence, with both the raw reads and consensus sequence labelled. Vague images below show a virus, coverage, and DNA.
32610
Reposted by Bede Constantinides
Nick Loman @pathogenomenick.bsky.social · 22/09/2026
Folks - if you're interested in how machine learning/AI can be used for microbial genomics, microbiome and metagenomics studies - please join us for a workshop in Birmingham on Thursday 8th October organised with CLIMB & NVIDIA. www.climb.ac.uk/blog/2026/09...
climb.ac.uk
Call for Applications: GPU-Accelerated Computing for Microbiome and Metagenomic Research · CLIMB
CLIMB is a cloud computing infrastructure for microbial bioinformatics — notebook servers, shared storage and reproducible workflows for genomics, metagenomics and pathogen surveillance researchers ac...
31920
Reposted by Bede Constantinides
Richard Sever @richardsever.bsky.social · 20/09/2026
"All corners of academia will soon be flooded with AI-generated research....Maths and CS go first because their entire output is symbol manipulation...Biology comes later not because biologists are cleverer but because the rate-limiting step is physical" giorgio.gilest.ro/the-weimar-o...
giorgio.gilest.ro
The Weimar of knowledge – la bottega
56640
Bede Constantinides @bede.im · 18/09/2026
As tasty as it is brimming with polyaromatic hydrocarbons 👀
020
Bede Constantinides @bede.im · 15/09/2026
Future HPRU methods day venue?
010
Reposted by Bede Constantinides
Will Matlock @wtmatlock.bsky.social · 09/09/2026
Does an AMR gene on a plasmid diversify differently to an AMR gene on a chromosome? We are looking for a PhD student to work on this question in Bath!
gw4biomed.ac.uk
Turning up the volume: plasmid copy number and gene duplications as drivers of antimicrobial resistance evolution - GW4 BioMed MRC DTP
Project Code IIAR27Ba Matlock Project Type Wet lab Research Theme Infection, Immunity, and Antimicrobial Resistance Project Summary Download Summary Plasmids are important drivers of antimicrobial res...
12627
Reposted by Bede Constantinides
Zamin Iqbal @zaminiqbal.bsky.social · 09/09/2026
More than a decade since Mykrobe first came out - v happy @martibartfast.bsky.social has developed a Go version, which means we could revive the old app, which always made it much more accessible. Same results as Mykrobe but faster.
0123
Bede Constantinides @bede.im · 08/09/2026
bsky.app/profile/jaco...
121
Reposted by Bede Constantinides
Stefan Neumann @neumannstefan.com · 08/09/2026
I wanted to submit a grant proposal this week and ended up not doing it precisely because of this. I have no idea how in a project that, say, lasts from 2027 to 2032, I can promise results that cannot be achieved by an LLM in 2031 with one person prompting.
1103
Reposted by Bede Constantinides
Ethan Mollick @emollick.bsky.social · 08/09/2026
This is a VERY big one. (And yes, the fights over academic credit and what happened in the race for the proof needs to be resolved, but it is still appears that this is a big one, if true.) openai.com/index/navier...
openai.com
On the Navier–Stokes Millennium Prize Problem
We’re sharing an AI-generated solution to the Navier–Stokes Millennium Prize Problem, including a writeup and a formal proof in Lean.
619731
Bede Constantinides @bede.im · 04/09/2026
As Pedro says, delegating to Opus seems necessary, as is keeping context/cwd clean of of e.g. virus names
110
Bede Constantinides @bede.im · 02/09/2026
Built by a friend :-) Thankfully I stayed upright and damage seems cosmetic
010
Bede Constantinides @bede.im · 01/09/2026
Reason #617 cyclists don't use cycle lanes
110
Reposted by Bede Constantinides
Richard Sever @richardsever.bsky.social · 27/08/2026
$3.7 billion spent on APCs in 2025. That's ~1000X more than it costs to run bioRxiv/medRxiv. Worth thinking about as we consider how to fund infrastructure... arxiv.org/pdf/2608.16322
arxiv.org
813378
Bede Constantinides @bede.im · 26/08/2026
You're absolutely right
000
Reposted by Bede Constantinides
Pathoplexus @pathoplexus.org · 26/08/2026
1/ 🦟 New on @Pathoplexus: Zika virus (ZIKV) is now supported! Share your ZIKV sequences as Open or Restricted-Use, just like our other pathogens. Read more: pathoplexus.org/news/2026-08... 📰
Screenshot of the Pathoplexus website homepage, where the new Zika virus button is circled in red
2167
Bede Constantinides @bede.im · 26/08/2026
"Load-bearing"
110
Reposted by Bede Constantinides
Philipp Markolin @philippmarkolin.eurosky.social · 18/08/2026
In 2026, there is no longer any lab leak hypothesis you can state consistently with the available evidence. Despite this, it has been the dominant media narrative for six years. Read my new data-driven investigation: www.protagonist-science.com/p/biased-med...
protagonist-science.com
Biased media coverage and the lab leak theory
Gain-of-function fearmongering and blame gets the glitzy NYT visual treatment as Republicans weaponize the lab leak theory again in attacks on Dr. Anthony Fauci
12342142
Bede Constantinides @bede.im · 20/08/2026
"A big wastewater scan that previously took about an hour just finished in six and a half minutes" @nickminor.bsky.social
000
Bede Constantinides @bede.im · 20/08/2026
"Smoke test"
030
Bede Constantinides @bede.im · 20/08/2026
Aside: even authentic human speech and writing has been influenced so much already. Perhaps it's Baader-Meinhof, but software-y folks in my circles have started saying things like "smoke test" and "materializing" out loud.
110
Reposted by Bede Constantinides
Neil Miles @neilsmiles.com · 17/08/2026
29th August 1985 - An excited Michael Fish gets to deploy the rarely seen red number as the temperature gets up to an incredible 26 degrees in London!
521655713
Bede Constantinides @bede.im · 18/08/2026
As requested, Deacon 0.17.0 adds BINSEQ CBQ support. CBQ is a binary sequence format by @noamteyssier.bsky.social that’s tiny on disk, fast to decode, and supports Ns. With CBQ, Deacon can filter sequences at >5 Gbp/s on a laptop (faster than uncompressed FASTA) github.com/bede/deacon/...
Comparison of Deacon host depletion throughput for cbq, cba, fasta, fastq and fastq.gz for simulated viral reads, with varying thread counts.
193
Bede Constantinides @bede.im · 15/08/2026
Awful news: "Tarling was hit by ​a non-race car driving in the opposite ​direction."
000
Reposted by Bede Constantinides
TCTMD @tctmd.bsky.social · 12/08/2026
David Taggart, Cardiovascular Surgeon and ‘Staunch Truthteller,’ Dies at 69
tctmd.com
David Taggart, Cardiovascular Surgeon and ‘Staunch Truthteller,’ Dies at 69
David Taggart, Cardiovascular Surgeon and ‘Staunch Truthteller,’ Dies at 69 mosbey@crf.org Wed, 08/12/2026 - 16:02 Yael L. Maxwell Known for his honesty and for igniting controversy, Taggart will be remembered for always putting patient interests first.
001
Bede Constantinides @bede.im · 08/08/2026
I get that airlines oversell flights, but should it not be illegal to pretend that check-in failed due to "technical issues" rather than running out of seats due to overselling? And then to advise checking in at airport when there's negligible chance of flying. Easyjet.
030
Reposted by Bede Constantinides
Kelly Sovacool @kelly.sova.cool · 06/08/2026
hot take: it is better not to have an image/graphic at all for your blog/talk/flyer/poster than to make one with genAI
0115
Reposted by Bede Constantinides
Aris Katzourakis @ariskatzourakis.bsky.social · 06/08/2026
HIV's hidden history: New study reveals much deeper roots stretching back millions of years sciencex.com/news/2026-07...
sciencex.com
HIV's hidden history: New study reveals much deeper roots stretching back millions of years
Researchers used to think that HIV and its relatives were pretty recent things, maybe only a few hundred years old. But a growing body of evidence now suggests their history is far more ancient, with ...
22112
Bede Constantinides @bede.im · 05/08/2026
I'd expect nothing else from a 90s film starring Mike Myers. To this end, I wonder what this new Ali G film will be like?
000
Bede Constantinides @bede.im · 05/08/2026
No! Had to google this reference 🥹
100
Bede Constantinides @bede.im · 05/08/2026
I need to watch Wayne's World!
130
Bede Constantinides @bede.im · 31/07/2026
This will make indexing k-mers in e.g. HPRC (human) and GTDB (bacteria) a breeze
030
Reposted by Bede Constantinides
Ragnar {Groot Koerkamp} @curiouscoding.nl · 31/07/2026
I've been working on Pandedup the past weeks: a tool to quickly dedup the k-mer content of a pangenome (read: HPRCv2). Just uploaded a k=64 SPSS (k-mer spectrum preserving string set) as well: zenodo.org/records/2172... This reduces the ~1.5Tbp input to just 10.6Gbp. github.com/RagnarGrootK...
github.com
GitHub - RagnarGrootKoerkamp/pandedup: Pandedup uses minimizers to build a quick & dirty k-mer spectrum (SPSS) from a .AGC pangenome.
Pandedup uses minimizers to build a quick & dirty k-mer spectrum (SPSS) from a .AGC pangenome. - RagnarGrootKoerkamp/pandedup
1104
Bede Constantinides @bede.im · 24/07/2026
Yes! To be published alongside a manuscript with a new parameter relevant to dehosting. Happy to share early though.
111
Reposted by Bede Constantinides
bioRxiv Genomics @biorxiv-genomic.bsky.social · 22/07/2026
HPRC2: A human pangenome reference with near-complete coverage of common genetic variation www.biorxiv.org/content/10.64898/20…
04024
Bede Constantinides @bede.im · 19/07/2026
Sympathies – it's already Outlook or bust (webmail) at University of Birmingham unfortunately.
010
Reposted by Bede Constantinides
Random Tiling Bot @randomtiling.bsky.social · 19/07/2026
This image shows a random tiling of a square created by the Random Tiling Bot
021
Reposted by Bede Constantinides
Ragnar {Groot Koerkamp} @curiouscoding.nl · 13/07/2026
Also, Deacon was the motivation for engineering the k-PHF Hash Set! bsky.app/profile/curi...
021
Bede Constantinides @bede.im · 14/07/2026
Aw! Thank you for making great stuff to build on! SimdMinimizers is just so useful, and it's great how even wasm gets to benefit from yours and @imartayan.bsky.social's simd work. I didn't even think to try disabling scalar mode for wasm at first 😅
010
Reposted by Bede Constantinides
Ragnar {Groot Koerkamp} @curiouscoding.nl · 09/07/2026
My secret 4th ALGO paper is out! We show a tight space lower on non-minimal k-perfect hash functions, generalize PtrHash into a non-minimal k-PHF, and then use it to develop a hash set implementation that is up to 1.6x faster than other hash sets! With Stefan {Hermann, Walzer} and Peter Sanders
2168