The anouncements this week were mostly focused on AI for diagnostics and automation, as well as some news on new sequencing assays and new RNA therapeutics.

This Week in TechBio 2024/02/05

• 20204/02/03 Twitter: An interesting application of diffusion models for designing regulatory DNA sequences. To the best of my knowledge it is the first “truly” use of modern generative models (besides the PGM-based ones) for DNA sequences.
• 2024/02/02 STAT: ARCH venture announces the creation of a new \$3000m funding round
• 2024/02/02 Fierce Biotech: Hologic gets FDA clearance for it AI diagnostic tool for predicting cervical cancer risks in pap smears.
• 2023/01/31 Nature: The current classification of cancers by their tissue of origin misses on the molecular origins and mechanisms that are shared by different cancers. This causes an issue because therapies known to wok on a given mechanisms have to be approved for each tissue of origin separately, leading to delays in access to therapies, as well as cancers with low number of patients never being targeted for clinical trials due to too small populations.
• 2024/01/31 BusinessWire: Genentech signs again with c3.ai for the use of its AI for manufacturing. This is a bit interesting because I’m pretty sure not a lot of folks who worked in machine learning have ever heard of that company, and they seem to be doing quite well for themselves.
• 2024/01/31 EndpointsNews: Sanofi really seems to be going all in on AI, going as far as making a lot of TikToks about how much they like AI. This is really reminding of Microsoft’s Steve Balmer famous “Developpers ! Developpers ! Developpers !”
• 2024/01/30 BusinessWire: Aqemia raises \$30m in a second round of its series A, putting the otal at \$60m. This round will be used to hire talent as well as push their internal discovery pipeline.
• 2024/01/30 STAT: Ultima genomics publicly announces its new sequencing machine that aims to reach the \$100 genome. Their machine will be more expensive and is aimed mostly at huge sequencing centers, but it could lead to the production of more genomics data, or as is showcased in the story the use of WGS instead of WES by default (Landau lab working on cell-free DNA). This technology has already been adapted for single-cell sequencing at the Broad in a 2022 Nature Biotech paper.
• 2024/01/29 Ascidian: First IND for a new class of RNA therapeutics, doing exon editing at the RNA level. Since RNA is all the hype for techbio, as it is a simple sequence problem, this new class of therapeutics can be a great application for machine learning models.
• 2024/01/29 Fierce Biotech: Samsung signs a deal to integrate Lunit’s chest X-Ray diagnosis AI into its devices. Lunit already had a deal with GE, and recently published a paper showing that its diagnosis tool outperformed radiologists.
• 2024/01/29 STAT+ and 2024/01/30 STAT: two interesting pieces on the different solutions considered by Medicaid to foot the bill for the new sickle cell treatment. Indeed as the price is in the \$2-3m range per patient, and there are quite a few patients (estimated of 100.000 in the US), the states do not yet know how to pay it. Solutions explored are an annuity-based solution, or a “outcome-based” solution where rebates will have to be put in place if certain objectives are not met (e.g. if the treatment does not work, pharma will have to reimburse the state). Some folks in the industry find this to be stiffing innovation, but if we are being honest, I’m not convinced that \$2m / person treatments are the innovation that society wants to encourage. The current cost of treating folks with sickle cell is \$3000m per year (mostly in hospital visits due to episodes of sever pain), which would be alleviated by the new curative treatment, but the cost of curing seems higher than the cost of treatment over a lifetime.
• 2024/01/29 Bloomberg: The DoJ is launching an investigation into how the use of AI, for recommending prescriptions drugs in EHR, can affect prescriptions patterns in a way that benefits pharma companies. In particular they want to know how they will be able to investigate potential kickbacks between EHR companies and pharma companies if the recommendations are done by black box algorithms.
• 2024/01/25 NEJM AI: Editorial on the new NEJM AI journal presenting their editorial stance. The explain that they are mostly looking for “clinical-grade” evaluations of AI in medicine, as well as trying to borrow some pages from th general Ai community with an interest un datasets and benchmarks.

I probably missed quite a few announcements, don’t hesitate to DM me @gama_search if you see anything missing or needing corrections.