Clare Bycroft retweeted
Introducing AlphaGenome: an AI model to help scientists better understand our DNA – the instruction manual for life 🧬
Researchers can now quickly predict what impact genetic changes could have - helping to generate new hypotheses and drive biological discoveries. ↓
Today the predictions from AlphaMissense (@GoogleDeepMind) have been integrated into @ensembl, @uniprot & the AlphaFold database. An important step in making predictions from machine-learning tools accessible to a wide audience. Great work @emblebi ! ebi.ac.uk/about/news/technol…
Clare Bycroft retweeted
We have received lots of interest in AlphaMissense, the @GoogleDeepmind model predicting the probability of a missense variant being pathogenic, since its release in September. Today we are updating the license terms for AlphaMissense predictions to also allow for commercial use.
Ever wondered what chatGPT's brain looks like? My brother made a very cool 3D visualisation and interactive tutorial dissecting the inner workings of a (small) large language model. Check it out!
Project #2: LLM Visualization
So I created a web-page to visualize a small LLM, of the sort that's behind ChatGPT. Rendered in 3D, it shows all the steps to run a single token inference. (link in bio)
Clare Bycroft retweeted
Less than a week to #wcpg2023, looking forward to seeing everyone there!
My lab is hiring postdocs to work on quantgen methods, EHR utilisation, disease heterogeneity and trajectories, if interested please DM, also happy to chat during the conference!
nacailab.com
Was a delight to present AlphaMissense at @CAGInews. Thanks very much to the organisers for an excellent conference! I learned a ton, and enjoyed all the lively discussions. May more of us 'catch CAGI' !
Clare Bycroft retweeted
This is a fantastic program that has had some superstars of human genetics pass through it. A chance to work with some incredible groups here in Oxford (not limited to those in WHG). Get those applications in!
We are accepting applications for 2024 entry to the four year Wellcome funded DPhil Programme in Genomic Medicine and Statistics, based here at WHG.
For more information, see well.ox.ac.uk/study/gms
Ever wondered which of those pesky missense mutations were likely to be pathogenic and which ones benign? Enter AlphaMissense! It’s not just a fancy AI model - it has real implications for human genetics. Juicy details now in Science scim.ag/49l
...prioritising variants for mutational scanning assays to precisely map the effect of variation on protein function...
Oh, and don't forget digging into the 60K different protein isoforms! Such a privilege to work with the brilliant @avsec, @juncheng, @joshbiology + others. Looking forward to seeing the creative ways others use this resource. dpmd.ai/alphamissense-blog
Clare Bycroft retweeted
The new AlphaMissense catalogue from @GoogleDeepMind can be used with the @ensembl Variant Effect Predictor to predict variant pathogenicity and therefore aid interpretation.
ebi.ac.uk/about/news/technol…
Clare Bycroft retweeted
We started the project by wondering how we could leverage the AlphaFold model (not just its predicted structures), for the pathogenicity prediction task. We were also inspired by methods like CADD/PrimateAI that train on population frequency data, and protein language models. 2/
Clare Bycroft retweeted
My lab @PioneerCampus is hiring!
We are looking for a postdoc interested in working on #singlecell and #spatial data to unravel the effects of chronic stress on neuronal gene expression and implications on psych disorders. DM enquiries welcome, pls RT!
tinyurl.com/34ck84dn
Really cool example of adapting DNA sequencing technology to be made more mobile and much cheaper, so can be used in many more contexts rnz.co.nz/national/programme…
So true! I can highly recommend trying to explain your science to 10yr-olds - harder than it looks, and a great way to improve your communication of science to anyone natureindex.com/news-blog/le…
Clare Bycroft retweeted
On Friday, we got the first delivery of samples from #UKBiobank. The delivery box contains 1470 samples. In a very short time, the content will give 2.6 M protein data points. That’s a lot of data from one small box, and in the end UKBB will receive 82.4 M protein data points.
Glad to see the X chromosome getting some love ♥️ nature.com/articles/s41593-0…












