Pharma AI Acquisitions, Data Origins

GSK is boosting its AI chops with a new deal worth up to $110 million with Relation.

The British Big Pharma and AI-focused biotech will work together in a new research collaboration focused on “generating large-scale human cellular perturbation data and deploying them into models,” according to a July 30 release.

Perturbation data are collections of biological measurements that track how systems respond to changes and are typically used in the early stages of drug development.

This is just the latest of many deals this year that Big Pharma did in an effort to start utilising the valuable data they sit on and bring the next blockbuster drug on the market. This all makes sense - as they go through pre-clinical and clinical studies, pharma companies generate massive amounts of deep data about a given disease, biological mechanism and a treatment method, and that includes a lot of failed attempts too. If there is a way to use those expensive data sets to flag early on in the drug development what is working and what not and to also generate reusable assets - that is worth a lot.

It also makes sense for AI model companies too. They need as much quality data as they can find to test and improve their models, and build a strong moat, fast. This means partnering with multiple companies and working on diverse indications in parallel.

That being said, it’s interesting to consider where the experimental and clinical data comes from. Some of it is done in-house, some at partner locations such as hospitals and CROs, but all of it is generated on devices and tools. In the old world, that data was kept locally and device manufacturers like Roche, Philips and Siemens Healthineers had no access to it. But with the promise of AI, MedTech business models are evolving from pure retail - buying machines or licences - to ones that have a usage component. Indeed, AI uses a lot of GPUs, which is expensive, and that cost needs to be covered by someone. There are some creative new ideas out there and I would be curious to see which ones succeed, but one thing is for certain - if AI is going to be any differentiator, then the business models will have to change, potentially paving the way for a new and much more robust data aggregator: the MedTech companies.

Healthcare Demand Drivers

This is the third point I want to address as part of my effort to explore the capacity constraints in the German healthcare system. Part one was looking at cost of care and part two from yesterday was about the origins of the physicians shortages in Germany. 200 years ago

Clinicians Shortages, Drivers

To continue the cost of care thread from yesterday, the second big capacity constraint is around clinicians. Here is a quick recipe for how to think about this problem: 1. First we take specialists such as radiologists and dermatologists (in-patient care) vs GPs (out-patient care). 2. Second we

Cost of Care Drivers

Last week I listed what I believe are the main factors contributing to capacity constraints in the healthcare system in Germany. Some of those are being already addressed in various ways, while others remain an open field (or shall we call it a blue ocean?): * Reimbursement incentives: Until recently, hospitals

Capacity Constraints

There are three main problems with the healthcare system in Germany: 1. Bad economics: cost of care is too high. This is driven by many factors, including reimbursement incentives, high manual work, various system inefficiencies and silos, and the medical field simply developing further and finding ever more cures for

Read more

Healthcare Demand Drivers

Clinicians Shortages, Drivers

Cost of Care Drivers

Capacity Constraints