Little bit of a “two for one” deal this week as I did not have time to correctly annotate all the news last week, so here comes the TechBio news of the past two weeks.

Overall there wasn’t a lot of announcements after JPM, I assume the folks are busy at Davos this week ? There were still a few interesting quotes from the CEO of Sanofi, Novartis and InSitro about how bullish or bearish pharma should be about the promises of GenAI (especially since it was the talk of the town at Davos), where CEOs seems unsure about which way to go. For my part I would agree with Daphne Koller’s take that while AI will be of tremendous use, the hard part is actually to get the biology and data part right, anyone can just grab a random model from huggingface and call it a day, but creating a clinical value from that requires much more work and is the actual hard part.

This Week in TechBio 2024/01/29

• 2024/01/26 Fierce Biotech: Deliberate AI’s digital health product for anxiety gets accepted into the ISTAND FDA program.
• 2024/01/25 Fierce Biotech: Merck is announcing a partnership with Variational AI, a company using generative AI for small molecules, where the prompt is a Target Product Profile.
• 2024/01/25 STAT: A paper recently published in the Lancet proposes a new AI model trained on EHR to predict risks of pancreatic cancer. It manages to identify a 3.5 larger risky population which should receive more screen for early detection. This is pretty cool and I guess many diagnostics company would want to work on that (even if getting access to a large cohort of EHRs will prove tricky), the business model may be hard to make with it though (who pays for being identified as “at risk”).
• 2024/01/25 TechCrunch: Innovation Endeavor, a VC firm founded by Eric Schmidt and focused on the applications of AI, raised fund V for \$630m
• 2024/01/24 Fierce Biotech: BenevolentAI gets a new CEO, by hiring a former Bayer exec. This is probably a great thing for BenevolentAI as it seems they lacked pharma experience, and approached biotech from a purely tech perspective.
• 2024/01/24 Fierce Biotech: GSK inks a deal with a linear DNA manufacturer Elegen, a few months after Moderna aquires OriCiro who does the same thing.
• 2024/01/24 Recursion: Altitude lab, an incubator specialized in techbios in Utah, has managed to reach \$120m in the total its portofolio companies raised. This incubator is also supported by Recursion.
• 2024/01/23 Fierce Pharma: Turquoise Health raised \$30m for its AI for price transparency in healthcare technology. This follows a rule that started its application in Jan 2021, mandating hospitals to post their rates, Turquoise aims to make the understandable.
• 2024/01/22 STAT: Researchers at the Whitehead recently published a paper in Nature, for doing lineage inference on HSCs, as some clonal issues are belived to be causal for anemia or leukemia. They rely on the fact that mitochondrial DNA accumulates mutations relativeley fast and can be used as a lineage barcode. They optimized a single-cell pipeline to capture this mtDNA as well as measuring the usual suspects (RNA and ATAC), and used tree algorithms (ggtree) to build a philogenic tree.
• 2024/01/19 STAT: Hospitals systems, such as Mayo Clinic and Mass General Brigham, are venturing into selling their internal data tools, as well as offer support for AI startup, in a bid to add new revenue streams. This sounds like a great opportunity to get data and early users for the AI diagnostics companies.
• 2024/01/19 JAMA: An opinion paper in JAMA warning against the dangers of using AI tools for digital pathology (using the example of kidney biopsies). The author’s position is that as AI tools get better and cheaper, pathologists will no longer need to spend a lot of time looking at histological data and will lose the knowledge of their analysis, as well as stop being able to discover new biology from them. I find this position to be rather silly and luddite, as it saves a lot of time and money to use these tools, and that wasting hours looking at the same images seems like a complete waste of time (the same way that cardiologists are bored at looking at holter EKGs). He complains that it makes pathology looks like just an input (image) output (diagnostic) problem, which is silly because it is exactly what it should be, and should be evaluated on, evidence-based medicine should also apply to diagnostics.
• 2024/01/18 Fierce Biotech: The Canadian non profit Conscience ran an interesting open science contest for AI based drug design targetting the LRRK2 WD40 repeat domain relevant for Parkinson’s Disease. While great for patients and society in general, I am a bit curious as to why pharma companies would join such a contest and publish their leads ?
• 2024/01/17 Yahoo! finance: Interview of Novartis’s CEO, where he mentions that while he doesn’t see AI change pharma in the next 5 years, he thinks that the changes done in the next 5 years will have a huge impact down the pipeline, but that the technology will need some time for its impact to be seen.
• 2024/01/17 STAT: a neat medical devices that tries to diagnoses the gravity of moles (benign vs malignant), it’s a bit better than dermatologists, but not super significant, still pretty cool.

• 2027/01/17 STAT+: STAT’s coverage of AI at JPM24, it covers al the deals announced last week, as well as shows that there is quite a lot of skepticism from folks not used to this technology. I have to say that I agree with them as just claiming “generative AI is better than regular old machine learning” is pure snake oil in my book (and I am an AI guy). Sanofi on the other hand seems really (maybe even too much ?) into that bandwagon.

  Sanofi’s chief executive, Paul Hudson, said the company has already built AI into the company’s daily operations. “More than 11,000 people across the company are using AI every day for better decision intelligence,” he said. “We actually are at a level now where we could ask the AI to tell us whether we should go and initiate a phase three (trial) with an important medicine.”

• 2024/01/16 Fierce Biotech: Rad AI does a partnership with Google to include MedPalm and other AI tools into its solution for radiology (diagnostics / report taking)
• 2024/01/16 Fierce Biotech: New data release from DepMap, which is a great source of data for target discovery for any lab.
• 2024/01/16 Whitelab Genomics: whitelab partners with another AAV company to provide target their AI target discovery toolkit.
• 2024/01/15 SIfted: CellVoyant raised \$7.4m in seed, Its technology is used to improve the growth of stem cell therapies (computer vision), to better monitor their differentiations.
• 2024/01/12 Fierce Biotech: FDA clears an AI diagnostics tool from MRI alone (cool because no need for a PET scan) for Alzheimer’s.

• 2024/01/07 STAT+: A very nice profile on what InSitro is doing with AI+Biology, it insists on why they decided to take so long to move their programs and did not have a lot of flashy annoucements; basically they are adults that want to do things right. There is a nice quote by Daphne Koller on wanting to become a engine and not a single asset company. Ewan Birney said good things, in particular praising their experiments design, and there is also this great quote:

  “These papers are good representative examples of what can be done today with cellular imaging and machine learning,” wrote Mark Murcko, who served as the founding CSO and is a director of two machine-learning-based drug startups: Relay Therapeutics and Dewpoint Therapeutics. He emphasized that “the real magic is NOT the machine learning.” There are clever people doing work in ML, he wrote, but that is becoming a commodity. The hard part, he noted, is the biology.

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