Excited to launch our AlphaGenome API goo.gle/3ZPUeFX along with the preprint goo.gle/45AkUyc describing and evaluating our latest DNA sequence model powering the API. Looking forward to seeing how scientists use it! @googledeepmind

I cannot recommend Davis’s group more highly! In addition to excellent research, he has been a great mentor!

📢 PostDoc opportunity in our Bioinformatics & Cellular Genomics lab at SVI! 🧬

You’d join a welcoming, supportive, and brilliant team.

Why not spend a few years in Melbourne and be part of something exciting?

Apply here: www.seek.com.au/job/84737876

#ScienceCareers #PostDoc #Bioinformatics

No serious person wants to be on a scholarship where the cohort has been hand picked by a bunch of trump lackies.

www.cbsnews.com/news/fulbrig...

Very happy to share that I will soon start a PhD at Cambridge University funded by the Harding Distinguished Scholar Fellowship, supervised by @mikeinouye.bsky.social and Angela Wood. I’ll work on prediction methodologies for health trajectories, using molecular data, and Electronic Health Records.

ALTMAN, on Musk:

“.. I wish he would just compete by building a better product. .. Probably his whole life is from a position of insecurity, I feel for the guy... I don't think he's, like, a happy person.”

@bloomberg.com

The FT reports OpenAI suspects DeepSeek of "a potential breach of intellectual property." As a columnist for NYT, which is suing OpenAI for copyright infringement, and the author of nine books OpenAI apparently used to train its model, I couldn't possibly comment. www.ft.com/content/a0df...

Molecular traits on the other hand are more neutral (because selection acts on the organism), thus are less confounded by population structure, which we believe further enhances ML prediction, because it allows the effect of distinct markers to be independent across populations.

Another interesting insight we make is the effect of ancestral selection on genomic predictions. We show how selection acts more strongly on complex traits, (like flowing time) causing variance in the phenotype to be collinear with population structure, which confounds genomic prediction

We find that ML applications best suit molecular traits. These traits tend to have simpler genetic architectures. We argue that with limited of observations, epistasis is more easily modelled when the variance in a trait is controlled by a few (e.g. 10) SNPs.

Very happy to share my first paper!

Here we take a look at different models for producing genomic prediction across a wide range of traits. We find that very specific conditions of where ML approaches can out preform linear regression.

Photos:
1. Media training
2. All of the Australians in attendance (I think we were the most represented nation per-capita)
3: Me and my friend Nikil outside the Heidelberg castle
4: Pat Hanrahan and some more friends

There were also opportunities to develop research adjacent skills inc. a media training workshop. I also presented an artistic representation of my research as apart of the “intercultural art project” that runs each year as apart of the conference. I highly recommend the conference!

It was amazing to also meet the other young researchers – every one there had something to teach me, and I learned about all sorts of fields from maths and compsci. I hope to stay in touch with the friends I made there!

Earlier this year, I was nominated and funded by unimelb to attend the Heidelberg Laureate forum in Germany. The conference connects 200 developing researchers from maths and compsci to award winning scientists inc. winners of the Fields Medal and Turing award.

It was an absolute blast!

Over 30 prominent scientists call for a ban on the creation of a "mirror cell"--a microbe made of molecules that are mirror images of their natural forms. It could cause a mind-boggling global disaster. Here's my story [gift link] 🧪https://nyti.ms/3OUCXp6