1Mby1M Virtual Accelerator AI Investor Forum: With Mo Islam, Partner at Threshold Ventures (Part 4)

Sramana Mitra: What is the situation vis-a-vis startup biotech companies that want to use a platform like Atomwise? Is Atomwise seeing startup biotech companies build on their platform?

Mo Islam: I don’t think it’s been as much about startups per se. Some companies want startups to build on top of them, but they’ve been targeting large pharma and more mature biotechs.

We have seen companies in the past that catered more to startups or various fields like drug discovery and fermentation, where people grow microbes for different use cases. We’ve seen a lot of those over the years, but they haven’t necessarily scaled up quite yet. However, that doesn’t mean the opportunity isn’t there.

Sramana Mitra: Is Verge built on atom wise?

Mo Islam: No, they are totally separate platforms. Verge is very vertically integrated, which excited us because they had novel access to data, especially human data. That was a big part of our bet.

We’ve spent a lot of time on vertical applications of AI. One of the key factors we look for, outside of the model itself, is the data used to train that model. In healthcare, drug discovery, and life sciences, access to novel data can make a significant difference. It’s different from scraping the internet for text, video, or audio data. Imagery, genomic data, and drug discovery data are not usually publicly available to a large degree.

Having novel access to data has been a big differentiator for some companies, both in drug discovery and in medical imaging and diagnostics. Companies like Imaging and VisAI started with novel data sets they had access to, which has made a big difference in terms of long-term defensibility.

Sramana Mitra: I’m going to ask you more about medical imaging in just a moment, but I want to provide a little bit of a synthesis for people who are listening.

If you are thinking about working in AI enabled drug discovery as a startup, and your primary unfair advantage as we call it, is on the biology side. You understand biology, you are coming into startups from the world of biology or medicine or where there is unique data, where there is unique knowledge of a drug discovery process, or coming from inside of pharmaceutical companies and you have some new idea, you kind of have to find a platform to work on.

It is very, very difficult and expensive to build the entire platform yourself. I think this discussion of the contrast between the strategy of Verge and the strategy of the platform company Atomwise are interesting in that context. If you are not coming from the technology startup or software kind of background and you’re trying to build more a drug discovery company based on your knowledge of biology, you do need to access software somehow.

This kind of software are very expensive to build. So platform as a service is going to emerge as an interesting model, and it’s something that you need to keep an eye on. Look for a platform on which you can apply your biology knowledge.

Do you want to comment on what I just said, Mo?

Mo Islam: No, I think that’s very fair. I would say we’ve seen a number of new companies emerge. For our particular fund, which is early-stage focused, we haven’t chased the big LLM companies like OpenAI and Anthropic. Instead, we’ve found interesting opportunities at the foundation model layer with companies building vertical-specific foundation models.

Some new companies are emerging in drug discovery, like Chi Discovery, which focuses on predicting and reprogramming interactions between biochemical molecules. Atomwise, from the last five years, is another example. We’re actively looking at new companies that could be interesting for startups to build upon, providing a significant leg up.

Even the bigger companies have shared phenomenal tools publicly. For example, Google DeepMind’s AlphaFold has been revolutionary in protein design. Demis at TED a couple of years ago mentioned that almost every biologist in an applicable area is using AlphaFold, which is amazing.

The toolset available now is much better than it was years ago, giving entrepreneurs a big advantage.

This segment is part 4 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: With Mo Islam, Partner at Threshold Ventures
1 2 3 4 5