Jesse Johnson
@jejomath.bsky.social
Software Engineer, AI for Pharma and Biotech
Weekly newsletter: https://scalingbiotech.substack.com/
Website: https://merelogic.net/
Weekly newsletter: https://scalingbiotech.substack.com/
Website: https://merelogic.net/
The announcement also claimed they're working to streamline access to this new data. Applying for CPRD, UK Biobank, etc. has always been a long and tedious process, so I'm less optimistic about that one. But given the sensitivity of the data, I think that's probably OK.
April 14, 2025 at 6:51 PM
The announcement also claimed they're working to streamline access to this new data. Applying for CPRD, UK Biobank, etc. has always been a long and tedious process, so I'm less optimistic about that one. But given the sensitivity of the data, I think that's probably OK.
Data like that could power the next generation of biomedical AI/ML models.
And it's the kind of thing that the NHS seems to be good at, so I'm cautiously optimistic.
And it's the kind of thing that the NHS seems to be good at, so I'm cautiously optimistic.
April 14, 2025 at 6:51 PM
Data like that could power the next generation of biomedical AI/ML models.
And it's the kind of thing that the NHS seems to be good at, so I'm cautiously optimistic.
And it's the kind of thing that the NHS seems to be good at, so I'm cautiously optimistic.
That's a huge number of genomes, but the NHS wants to do better.
So the new initiative seems to be aimed at getting more genomic and other biomedical data for the tens of millions of patients who are represented in the healthcare datasets but not UK Biobank.
So the new initiative seems to be aimed at getting more genomic and other biomedical data for the tens of millions of patients who are represented in the healthcare datasets but not UK Biobank.
April 14, 2025 at 6:51 PM
That's a huge number of genomes, but the NHS wants to do better.
So the new initiative seems to be aimed at getting more genomic and other biomedical data for the tens of millions of patients who are represented in the healthcare datasets but not UK Biobank.
So the new initiative seems to be aimed at getting more genomic and other biomedical data for the tens of millions of patients who are represented in the healthcare datasets but not UK Biobank.
The UK Biobank, on the other hand, is all about these biomedical signals, particularly whole genome sequences. These are linked to the healthcare data, but it only has about 500K volunteer participants.
April 14, 2025 at 6:51 PM
The UK Biobank, on the other hand, is all about these biomedical signals, particularly whole genome sequences. These are linked to the healthcare data, but it only has about 500K volunteer participants.
These are huge datasets but their use for drug discovery is limited because they aren't linked to biomedical signals like genomics.
April 14, 2025 at 6:51 PM
These are huge datasets but their use for drug discovery is limited because they aren't linked to biomedical signals like genomics.
In 2012, the NHS launched CPRD, a longitudinal dataset of de-identified primary care medical records for tens of millions of patients. And the NHS provides lots of other datasets covering in-patient care and other parts of healthcare delivery.
April 14, 2025 at 6:51 PM
In 2012, the NHS launched CPRD, a longitudinal dataset of de-identified primary care medical records for tens of millions of patients. And the NHS provides lots of other datasets covering in-patient care and other parts of healthcare delivery.
It seems more like it was the superheroes who were struggling with a society in which they were no longer needed and thus had no way to assert their dominance over others.
April 13, 2025 at 11:33 AM
It seems more like it was the superheroes who were struggling with a society in which they were no longer needed and thus had no way to assert their dominance over others.
I would think the APIs are the most likely thing Siemens would address, since they can probably do it without impacting user experience, and it would be important for digital twin integration.
Not that I'm saying you should be optimistic. Just that it's not impossible...
Not that I'm saying you should be optimistic. Just that it's not impossible...
April 12, 2025 at 10:41 PM
I would think the APIs are the most likely thing Siemens would address, since they can probably do it without impacting user experience, and it would be important for digital twin integration.
Not that I'm saying you should be optimistic. Just that it's not impossible...
Not that I'm saying you should be optimistic. Just that it's not impossible...
It might be interesting to look at Siemens' 2021 acquisition of FORAN software, which was also motivated by digital twins, to see what happened there. But I don't know enough about ship design software to do that analysis.
April 12, 2025 at 8:25 PM
It might be interesting to look at Siemens' 2021 acquisition of FORAN software, which was also motivated by digital twins, to see what happened there. But I don't know enough about ship design software to do that analysis.
Both Siemens' and Dotmatics' press releases mention digital twins and AI capabilities as the main motivation. This can all happen through the analytics layer, so it's very possible that Siemens will take a similar hands-off approach.
April 12, 2025 at 8:25 PM
Both Siemens' and Dotmatics' press releases mention digital twins and AI capabilities as the main motivation. This can all happen through the analytics layer, so it's very possible that Siemens will take a similar hands-off approach.
Given the price tag of the acquisition, this strategy seems to have worked for Dotmatics. But will Siemens be looking for deeper integration?
April 12, 2025 at 8:25 PM
Given the price tag of the acquisition, this strategy seems to have worked for Dotmatics. But will Siemens be looking for deeper integration?
Insightful science, the company that acquired Dotmatics, GraphPad, SnapGene and a number of other tools before changing its name to Dotmatics, mostly left the products themselves alone. The parent company focused on integrating them through the analytics layer (Luma).
April 12, 2025 at 8:25 PM
Insightful science, the company that acquired Dotmatics, GraphPad, SnapGene and a number of other tools before changing its name to Dotmatics, mostly left the products themselves alone. The parent company focused on integrating them through the analytics layer (Luma).
Seems like a real "Never let a good crisis go to waste" moment.
Are you worried that optimizing for AI will eventually turn into the kind of dynamics we see with SEO today?
Are you worried that optimizing for AI will eventually turn into the kind of dynamics we see with SEO today?
March 18, 2025 at 4:20 PM
Seems like a real "Never let a good crisis go to waste" moment.
Are you worried that optimizing for AI will eventually turn into the kind of dynamics we see with SEO today?
Are you worried that optimizing for AI will eventually turn into the kind of dynamics we see with SEO today?
There's this weird dynamic I've seen in multiple contexts where developers assume they can't do anything about upstream processes, so they try to fix all the data downstream.
Plus downstream transformations are closer to analysis so they get more social status.
Plus downstream transformations are closer to analysis so they get more social status.
January 6, 2025 at 10:33 PM
There's this weird dynamic I've seen in multiple contexts where developers assume they can't do anything about upstream processes, so they try to fix all the data downstream.
Plus downstream transformations are closer to analysis so they get more social status.
Plus downstream transformations are closer to analysis so they get more social status.
So what does this have to do with AI/LLMs?
Well, one thing that even today's AI/LLMs do is make complex, esoteric information accessible by summarizing and digesting it to answer specific questions.
Yes, they're wrong a lot, but they're right enough that we keep using them.
Well, one thing that even today's AI/LLMs do is make complex, esoteric information accessible by summarizing and digesting it to answer specific questions.
Yes, they're wrong a lot, but they're right enough that we keep using them.
January 5, 2025 at 2:00 AM
So what does this have to do with AI/LLMs?
Well, one thing that even today's AI/LLMs do is make complex, esoteric information accessible by summarizing and digesting it to answer specific questions.
Yes, they're wrong a lot, but they're right enough that we keep using them.
Well, one thing that even today's AI/LLMs do is make complex, esoteric information accessible by summarizing and digesting it to answer specific questions.
Yes, they're wrong a lot, but they're right enough that we keep using them.
If you're choosing between a restaurant that you know is open (because their hours are on Yelp or Google maps) and one where you're not sure, you're probably going to pick the first one.
Accessibility of this information made availability a competitive advantage, and eventually an expectation.
Accessibility of this information made availability a competitive advantage, and eventually an expectation.
January 5, 2025 at 2:00 AM
If you're choosing between a restaurant that you know is open (because their hours are on Yelp or Google maps) and one where you're not sure, you're probably going to pick the first one.
Accessibility of this information made availability a competitive advantage, and eventually an expectation.
Accessibility of this information made availability a competitive advantage, and eventually an expectation.
Then Yelp, and later Google Maps, made it really easy to find this information - for the restaurants that published it.
Customers could now compare hours side by side.
Customers could now compare hours side by side.
January 5, 2025 at 2:00 AM
Then Yelp, and later Google Maps, made it really easy to find this information - for the restaurants that published it.
Customers could now compare hours side by side.
Customers could now compare hours side by side.
Even when the information was available, it wasn't accessible.
It was easier to call the restaurant than to look it up, so there wasn't an incentive or expectation for restaurants to publish it.
It was easier to call the restaurant than to look it up, so there wasn't an incentive or expectation for restaurants to publish it.
January 5, 2025 at 2:00 AM
Even when the information was available, it wasn't accessible.
It was easier to call the restaurant than to look it up, so there wasn't an incentive or expectation for restaurants to publish it.
It was easier to call the restaurant than to look it up, so there wasn't an incentive or expectation for restaurants to publish it.
In the early days of the internet, some restaurants put their hours of operation on their websites. But most of them didn't, and even the ones that did wouldn't necessarily update them when they changed.
It was too much of a pain to look it up, so most people didn't.
It was too much of a pain to look it up, so most people didn't.
January 5, 2025 at 2:00 AM
In the early days of the internet, some restaurants put their hours of operation on their websites. But most of them didn't, and even the ones that did wouldn't necessarily update them when they changed.
It was too much of a pain to look it up, so most people didn't.
It was too much of a pain to look it up, so most people didn't.
The major dynamic I'm thinking about right now is how information accessibility changes incentives and expectations for the kinds of information that companies publish.
I'll explain with an example.
I'll explain with an example.
January 5, 2025 at 2:00 AM
The major dynamic I'm thinking about right now is how information accessibility changes incentives and expectations for the kinds of information that companies publish.
I'll explain with an example.
I'll explain with an example.