Next up: ‘From Axons to AI: A Biologist’s Journey’, by Marc Tessier-Lavigne, CEO of Xaira, interviewed by Matthew Herper. #STATBreakthrough

Side note: I helped with Theo Baker’s (his book came out today!) reporting on image problems in MTLs paper, eventually causing him to step down as the Stanford University president. Let’s see if this will be discussed today. stanforddaily.com/2022/11/29/ #STATBreakthrough

November 29, 2022 - The Stanfo...

MTL: 90% of drugs fail in clinical trial. AI can make a big difference in improving clinical trial, both on designing better drugs as well as matching patients to the targets. AI models can help in particular with complex cellular pathways, how to make cell healthy again. #STATBreakthrough

MTL: We can now make new molecules from scratch using AI. AI can open new doors. MH: you resigned from Stanford. Some images might have been manipulated but you were not find guilty of knowing this. Did you adequately correctly the record? #STATBreakthrough

MTL: I have been cleared of misconduct. I always trusted my staff but I have learned that you need to balance that with attention to potential misconduct. #STATBreakthrough

MTL: I sent several emails asking for corrections were lost by the journal, and I could have done better following up. MH. Theo Baker’s book came out today. I will read some paragraphs. #STATBreakthrough

MH reads some paragraphs from Theo Baker’s book, about the board of trustees with a not very flattering description about MTL’s conduct by anonymous persons. MTL: I can be firm but I try to be fair. I need to be mindful that I can come across as strong. #STATBreakthrough

MH: how prepared are you to deal with the public trusting you r company? MTL (sounding relic that we moved on to a different topic): we have to earn this trust. China is making stunning progress in new molecule development. We are more guarded in releasing new information. #STATBreakthrough

MH is you could go back to 2022, what would you tell yourself? MTL: I would like to go back to 2015 and pressed harder to correct those papers. I have been corporative with the investigation panel. MH. Would you have skipped the whole Stanford thing? MTL: No, I am proud of it. #STATBreakthrough

* If you could go back

Amazing view of the Bay Bridge from the Commonwealth Club in San Francisco, where we are having lunch break. #STATBreakthrough

Back from lunch with ‘Gene Therapy's Next Horizon: The Womb’ with Tippi Mackenzie @ UCSF, interviewed by Megan Molteni. Can genetic editing and fetal surgery become standard of care? #STATBreakthrough

TM tells about how she got interested in fetal surgery and is now delivery mole therapy to a fetus developing in the womb. This could be anatomical problems or genetic issues, which we can now treat with gene therapy or stem cell transplants. #STATBreakthrough

TM: We had to wait until those techniques were developed in adult. Some diseases develop during the fetal stage so treating them after birth has only limited effects. MM: Do patients really want this? TM: Not all, some terminate pregnancy. You have 2 patients: mother and fetus. #STATBreakthrough

TM: There are all kinds of ethical issues, such as unknown risks and complications. There has to be a big clinical improvement such as fixing a CNS disease. We have to walk a fine line between risks and benefits. We need more biomarkers to monitor fetal responses. #STATBreakthrough

TM: we do a lot of basic research but need more awareness to discuss issues that gene therapy could address. MM: where are we going to be 5 years down the road? TM: hard to answer! we need to learn by doing, experience the bottlenecks and move through them. #STATBreakthrough

Next: The Drug Discovery Blueprint: Synchronizing Physics, Intelligence, and Federated Learning * Aliza Apple, Eli Lilly * Jonah Cool, Anthropic * Aleksey Gerasyuto, Schrödinger * Jesse McQuarters, moderator Single scientists can now work with teams of AI tools. #STATBreakthrough

AA: we launched TuneLab, derived from Lilly’s research and make this available to biotech companies. investor.lilly.com/news-release... JC: a scientist should understand what the agent did. You should be able to reproduce it. The models should look it up, not make it up. #STATBreakthrough

AG: patient safety is important. Drug molecules might already be optimized for efficacy and it would be time consuming to change the drug molecule at a late stage to incorporate patient safety. AA: Not just providing predictions, but also confidence intervals. #STATBreakthrough

AA: We can now get to clinical compounds in two years. It does not take the human scientist out of the equation. JC: if we need to get one thing right with AI models, it is health and medicine. #STATBreakthrough

New panel: Explorers of AI’s frontier. * Lauren Kopsick, Healthcare Navigation project * Vincent Lou, Permanente Medical Group * Mark Sendak, Vega Health * Casey Ross at STAT, moderator LK: there is still some friction. We don’t always know why AI takes certain decisions. #STATBreakthrough

MS: a void of capacity to interpret technologies. It is moving incredibly fast. CR: are patients involved enough? VL: with the pace we are moving, patients need to be involved. AI might be a wedge that drives patient/doctors apart. That might not be archieve better health. #STATBreakthrough

LK: Doctors are very frustrated with all of this. The process is breaking down with AI. I love AI, but staff, doctors, and patients need more education. Q: how do you educate patients? VL: Ambient is Kaiser’s AI scribe tool to record conversations; saves MDs lots of time. #STATBreakthrough

VL: Human/AI handoff. 1) AI is great but can get it wrong. 2) Protect our patients’ data. 3) Always allow feedback. We need to keep on training the AI. MS: Organizations need to implement new technology but trust is needed too. #STATBreakthrough

Next: Lison Joseph interviews Christine Lemke at Evidation. CL: I have autoimmune disease most often diagnosed in white men of a certain age. It might take years to diagnose a disease if you don’t fit the standard presentation. If we could all monitor ourselves earlier diagnosis. #STATBreakthrough

CL: Wearable devices can send signals before a flare up, create large amounts of data. AI can eat into every layer of a company. Lots of people think that AI is going to take their jobs. As a CEO, keeping people motivated is one of my concerns. #STATBreakthrough

Next: The wearables revolution, with: Ricky Bloomfield of Oura Alex Vannoni of WHOOP Moderated by Katherine MacPhail of STAT, who asks the audience who is wearing a wearable - most of us! #STATBreakthrough

AV: if we add blood work to wearables, our members can pick up lots of additional biomarkers. It is opt in, 4 times per year KMP: isn’t that creating too much data and creating anxiety? AV: it is to create a base line and act on it. #STATBreakthrough

RB: lots of people do not trust the traditional health system. If they avoid regular screening, this might be a way to get people treated. AV: we see this a complimentary to the current system. Human experts are not going away. It is a system that builds, rather than competes. #STATBreakthrough

RB: If we want to improve people’s health, we need to measure data. AV: we can reach new cohorts, new partnership with Medicare. www.whoop.com/us/en/press-... Great sock game going on! #STATBreakthrough

AV: the FDA and other regulators are more open now to use wearables as medical devices (eg, blood pressure, arythmia) and we are very open to regulation. RB: new FDA guidance is a welcome update, Wearable can not diagnose but can monitor and connect to health provider. #STATBreakthrough

Coffee break time! #STATBreakthrough

Next: Anne Wojcicki @ 23andMe with ‘Anne Wojcicki won’t give up’, moderated by @[email protected] MH: The company has been through a lot, FDA regulation, and recently bankruptcy. You also went through lots of family loss. #STATBreakthrough

AW: we wanted to give people direct access to their DNA, so we started a direct-to-consumer company. How can we combine this with medical records? You can now upload your medical record. The combination plus AI tools will help with diagnosis. #STATBreakthrough

AW: self-reporting is not always reliable (think Alzheimer’s), so combining medical records with genetics will fill some gaps. Can we predict diseases in the future? The bankruptcy was not my choice, I I felt awful. But employees were ok: they focused on the mission. #STATBreakthrough

AW: so I decided to make a bid to buy the company with my foundation. MH: a lot of people were very worried about privacy when the company went bankrupt and when there was a data breach. AW: we communicated with millions of our customers - lots of people wanted to sue us. #STATBreakthrough I

AW: That was devastating for us. We tried so hard to earn customer trust. We ask people to opt in. We let people delete their account - and 2 million people did that. It was painful, but we honored that. #STATBreakthrough

Audience question: there likely will be a new data breach. How do we safeguard data? Insurance companies might get their hands on it, and raise premiums. AW: we now mandate two-factor authorization and log you out quicker. There is always a risk. #STATBreakthrough

Next: The Rare Disease AI Inflection Point, with Arpit Davé at BioMarin. Moderated by Rick Berke. AD: BioMarin focuses on rare diseases. Our North Star is to bring therapies to patients who need it the most. AI can help with this. #STATBreakthrough

AD: many rare diseases do not have their own ICD (International Classification of Diseases) code or are binned together, making it hard to analyze data. LLMs are good at reading human language so can look in more detail. #STATBreakthrough