FDA is supporting blockchain tech & Mayo is leveraging federated learning

Blockchain in healthcare updates

The FDA’s tech modernization plan highlights support for blockchain

The plan mentions blockchain technology as one of the technologies the FDA needs to support to modernize their infrastructure. There are a few mentions of blockchain, but track and trace was the only use case concretely highlighted. There was one mention of blockchain tech I found particularly interesting in the section outlining the FDA’s plans to develop tools around specific regulatory “use cases:”

Provide a “how to” roadmap to address key tasks such as utilizing cloud technologies for FDA purposes, appropriate data sharing with FDA, and application of new technologies like blockchain to FDA challenges; and…

I’m not sure what the FDA will release, but it could be very influential for the industry as they’ll likely need to make choices about what blockchain to use, how much data goes on chain, how identities are managed, etc.

This is the modernization plan that Dr. Amy Abernethy, CIO for the FDA, teased a few months back. I would have liked to see use cases other than track and trace highlighted but regardless it is nice to see regulators enthusiastically supporting emerging technologies like blockchain.

Molecule’s first research project is live and being funded

Molecule is a startup creating a "software platform to accelerate innovation in the pharmaceutical industry." The first part of this is using novel token mechanisms to enable crowdfunding, distribute intellectual property, and incentivize collaboration. On Thursday the first project on Molecule, a clinical trial in Canada testing the effects of Psilocybin, went live and raised ~$2,000 in a few hours. You can’t invest in this effort, only donate to it, but there’s now a way for anyone in the world to directly contribute to this research. That’s pretty cool and has a lot of potential.

The model they are using is more complex than simple donations. I won’t try to explain it here, but even as someone who spends an inordinate amount of time in this space I found it a bit hard to grok. Translating complex concepts and creating a great user experience for mainstream users will be an important challenge going forward.

On Lympo’s new token model

Lympo is a blockchain and healthcare startup that has created a fitness app that rewards users with tokens. They secured a spot in Samsung’s crypto-wallet, ensuring a desirable distribution channel for their application. But alas, even a relatively popular app with solid partnerships hasn’t been able to sustain the price of the token issued by Lympo. In a bid to change that they announced that they would be using 10% of their profits to buy back their tokens on the open market.

This general model of using profits to buy back tokens was pioneered to much success by crypto-exchange Binance. Buying tokens in this way increases demand and reduces supply; all other things being constant this theoretically leads to an appreciating price. That’s desirable for a business as it drives interest in your project and makes for happy community members who are often also your best evangelists. Moreover, a token that is stable or appreciating is necessary for any “incentive mechanism” to work beyond the margins. There has been much talk of “incentivizing people” to do things but in the long term people need to expect that tokens will be worth something for them to care enough take action.

There are some issues to be resolved around this the (economics, legality, transparency) and ultimately the market will decide how to interpret this move. But as I’ve said multiple times I welcome and expect experimentation intended to create value.

What I’m reading this weekend

The Mayo Clinic’s novel data privacy model leverages federated learning

The Mayo Clinic announced this week at JPM20 that the first venture of their platform would be the “Clinical Data Analytics Platform” and they’ve partnered with a startup called nference for that venture. They also announced their “data privacy model,” and notably they are making federated learning a core part of the privacy model for their platform. By doing so external partners can query the Mayo Clinic’s extensive data but all that gets shared with those parties is the answers to their queries. None of the underlying data is shared. I was surprised at how little press coverage this got, it represents a sharp break from the way that health systems had previously worked with tech companies, and a new strategic direction for health systems.

How this differs from other federated learning deployments: Federated learning is usually deployed in a network, where multiple parties join together to collaboratively train an algorithm. This is the case for MELLODDY or KCL’s federated learning initiative. In these networks there needs to be some kind of orchestration of actions that decides what algorithm to train and what data to use. It could be problematic for one party to unilaterally decide this and thus the aforementioned federated learning deployments use a blockchain to decentralize this orchestration process.

In contrast Mayo is the only data provider here and by definition orchestration is centralized, so there isn’t a need for a blockchain. Federated learning like this is still valuable but is most compelling in a network setting. This announcement and Mayo’s platform aspirations makes one wonder whether they intend to add other members and create a network.

Facebook is sponsoring research into privacy preserving tec, including zero-knowledge proofs and federated learning

Epic warms customers it will stop supporting Google Cloud

Keep in mind Google’s aspirations to create a new EHR.

Biomarkers and Receptor Targeted Therapies Reduce Clinical Trial Risk in Non–Small-Cell Lung Cancer

Biomarker targeted therapies (62%) and receptor targeted therapies (31%) were found to have the highest success rates. The risk-adjusted cost for NSCLC [non-small-cell lung cancer] clinical drug development was calculated to be U.S. $1.89 billion.

Comparison of Machine Learning Methods With Traditional Models for Use of Administrative Claims With Electronic Medical Records to Predict Heart Failure Outcomes | Cardiology | JAMA Network Open

…machine learning methods offered only limited improvement over logistic regression in predicting key outcomes in heart failure based on administrative claims. Inclusion of additional predictors from electronic medical records improved prediction for mortality, heart failure hospitalization, and loss in home days but not for high cost.

Former CFTC Chair to setup a nonprofit to promote digital dollar

As other countries have rapidly progressed towards creating their own digital currencies the US, which has the most to lose, has not done much. Giancarlo was the chairman of an important financial regulatory body recently and has been an outspoken advocate for a “digital dollar,” basically turning dollars into a token on a blockchain. This nonprofit will help advocate that case.

To that end, Brian Armstrong, CEO of Coinbase, said there are “conversations taking place about how the dollar can be digitized”:

The U.S. is playing a bit of catch up now, and active discussions are taking place about how the dollar can be digitized. CENTRE, with its USD Coin, may be the solution that U.S. turns to, or the Fed may try to implement their own digitized dollar using blockchain. I think we will then see basket digital currencies come out, either by a consortium like Libra or CENTRE, or possibly the IMF itself.

Federal Health IT Strategic Plan

23andMe has sold the rights to develop a drug based on its users’ DNA


If you found this newsletter valuable then you can click the button below to sign up for free.

If you’re an existing reader I would deeply appreciate it if you share this with people who would find it a valuable resource. You can share this piece on LinkedIn here, and there is a button for sharing on Twitter below. You can also “like” this newsletter by clicking the heart just below this, which helps me get visibility on SubStack.

My content is free, but if you would like to support me, you can do so on Patreon.

Feel free to reply to this email with comments, questions, or feedback. I host a blockchain and healthcare Telegram group that you can join. You can also find me on Twitter here.