Announcements
I’m officially deprecating my ‘Interesting Links’ posting. I feel like I prefer something richer and more information dense, and ‘roundup’ fits that need better. I’d also like these sorts of posts to function as ways to announce stuff about this blog, alongside some interesting links + jobs (more on that later). Also removing the paywall, because it wasn’t particularly useful as a funnel anyway.
So, announcements!
One, I’m ‘hosting’ (reserved 4 tables for free) a bio-ML event at a bar in Williamsburg on Wednesday, Feb 5th from 7:30pm-10:30pm. Come by and hang out! Here’s the Partiful link w/ more information.
Two, I spent the winter holidays learning a bunch of graphic design, and got really into poster-making. Specifically, biology/biopunk posters, which is a niche that surprisingly few people are creating for. Early iterations of them had pretty good reception, so I’ve decided to sell them, you can find the shop here. Here’s one of my favorites:
I’d like to make tons more of these, so reach out if you have a fun idea for something! I’ve collaborated with two companies to turn their research into a poster (the one above, and another on my shop here) and would like to do it with others.
Three, and finally, these roundups will now include a ‘Job’ section! Lots of non-bio SWE’s/ML/product people read this blog and ask me where they should be applying to if they want to get into biology. I have no idea usually (other than, of course, Dyno Therapeutics). If you’re a founder or very happy employee at a biotech (stealth or otherwise), feel free to DM me with the role and I’ll add it here! Ideally will be focused on engineering/computational/product roles.
Links
The branch of TF’s they focused on were redesigning two Yamanaka Factors (SOX2 and KLF), which are a total set of 4 TF’s meant for converting normal cells to stem cells. You can do a partial ‘conversion’ of cells for the purposes of life extension, which I’ve written about before.
TF’s have a pretty high rate of intrinsically disordered regions, hence why they went with pure sequence over structure.
Very few details on what exactly they did.
Quote from the article: OpenAI’s new model, called GPT-4b micro, was trained to suggest ways to re-engineer the protein factors to increase their function. According to OpenAI, researchers used the model’s suggestions to change two of the Yamanaka factors to be more than 50 times as effective—at least according to some preliminary measures.
What does ‘50 times more effective’ mean? Conversion rate? Compared to what baseline? Vague! The below picture is the only bit that somewhat pokes at what happened, potentially implying that the improved two TF’s are better compared to the standard set of 4 Yamanaka Factors. But there’s still a lot of factors to call out. Did these cells develop cancer at a higher rate? What were they reprogrammed to? Did their epigenetic age still decrease? Riffs on the Yamanaka Factors are quite common, but it seems like relatively few of those riffs have stuck around.
Hopefully more details come out soon. Cool that OpenAI is getting in on biology though!
No real changes to the original preprint, which was released back in June 2024. I’ve written about the paper before here, specifically about the protein they redesigned (GFP).
Colossal Biosciences (the ‘bring wooly mammoths back’ startup) raises $200M at $10.2B valuation.
There seem to be two camps here.
I think both sides are right. I hope Colossal succeeds and I want more crazy companies! But I also think $10.2B is a bit much for a company that hasn’t shown positive results…ever. If you’re curious, Asimov Press wrote an excellent overview of ex-vivo technology back in October 2024, and specifically touched on Colossal’s ambitions (which are indeed lofty).
Spoiler alert: he is very optimistic! We’re definitively entering the realm of ‘on demand protein binders’ for at least some targets. There are concerns with the developability of these binders (not aggregating, etc), but that’s a far easier problem than coming up with the binders in the first place.
If curious, they use RFDiffusion for generation. Binding results held in both in-vitro and in-vivo settings, successfully protecting a mouse from snake venom.
Jobs
As mentioned at the start of this, I will now start advertising jobs! Will try to keep it minimal and only for places I think are working on cool things. Please reach out to me if you are hiring and would like to be a part of this! Will only include you here once per roundup, unless you reach back out to me again.
Senior/Staff Software Engineer - Full Stack (Dyno Therapeutics)
I work here, helping out on protein engineering efforts to solve the ‘delivery problem’ in gene therapy. Obviously I have a biased opinion, but Dyno is one of the very few AI-biotech companies with clear product-market-fit. So betting on us isn’t a bad idea!
Founding Software Engineer (Tamarind Bio)
From the founder: Thousands of scientists from top biotechs and top 20 global pharma companies use our software to discover and optimize protein drugs. We are profitable, well-funded from investors including Y Combinator, and hiring to keep up with 20+% monthly growth. Our founding engineer position involves scaling inference to thousands of GPUs, deploying the state of the art AI models for customer use in practical drug discovery, and automated fine-tuning pipelines.
Product Engineer, ML Infrastructure Engineer, Software Engineer, and Life Science Specialist (Convoke Bio)
From the founder: Convoke is building the autonomy software stack for drug development. Our biotech and pharma customers use our infrastructure to automatically source and integrate information that feeds their most critical decisions. We're hiring ML and full-stack engineers to quickly build out our product suite, working closely with our customers. Our founding team has deep life science experience and has previously built several large vertical businesses and high growth technology startups. We are well capitalized, with the backing of top tier investors. Our team is based in South San Francisco.
My personal take: Alex Telford, the CEO of the company, has one of the most informative and deeply researched biology blogs I’ve ever read. I feel like it’d be quite challenging to read some of his articles and not come away deeply optimistic on whatever the writer of such material works on. You should apply!
Senior Data Engineer (Pattern Bio)
From the JD: At Pattern Bio, we are creating next-generation cancer therapies at the intersection of synthetic biology and machine learning, with data as the cornerstone of our platform. Our mission is to transform disease treatment, starting with cancer, by leveraging our innovative biomolecular computing technology. By utilizing multi-input, molecular-level computation within individual cells, we aim to deliver curative therapies where the traditional one drug-one target approach has fallen short. Pattern Bio is assembling a world-class team to bring two decades of advancements in DNA computing and machine learning into the clinic
I do hope you get more postings for bioinformatics or comp bio roles in the future . Even the most exciting, cutting edge companies can't all be ML Engineering and Ops right? Maybe it's just observation bias.