AI drug discovery company Talus Bioscience raises $11.2M

The company's platform, MARMOT, can identify molecules to block disease-causing proteins

Talus Bioscience was founded to discover molecules that target transcription factors (TFs), or proteins that regulate the genome by binding to DNA and orchestrating which genes are turned on or off at any given time. When TFs go awry, either through mutation, deletion, or hyperactivity, they can then switch on cellular processes that lead to different diseases, such as cancer, fibrosis, and inflammatory conditions.

Yet, TFs have historically been considered “undruggable” despite being some of the most promising targets in oncology, Alex Federation, Ph.D. CEO and co-founder of Talus Bisoscience, told VatorNews.

"These proteins have been considered 'undruggable' because traditional drug discovery approaches that remove them from their native environment inside the cell have failed. They only fold and function properly inside the cell. For similar reasons, we can’t observe the three-dimensional structure of these proteins, so other advances in AI for drug discovery, such as AlphaFold, aren’t useful for these targets either," he explained.

Talus Bioscience's platform, called Multiplexed Assays for the Rational Modulation Of Transcription Factors (MARMOT), fixes this by combining automated biology, scalable proteomics, and ML/AI so it can identify molecules to block disease-causing TF activity in this native cellular environment. Now the company is looking to advance the platform to a number of therapeutic disease, which it will be able to do thanks to an $11.2 million venture funding round announced on Tuesday.

The seed+ round was led by Two Bear Capital, with participation from WRF Capital, NFX, YC Continuity Fund, Funders Club VC, and BoxOne Ventures. To date, Talus Bio has raised six non-dilutive grant awards for over $7.3 million combined, along with $19.7 million in venture funding, including this latest round.

The MARMOT platform allows Talus to measure the activity of a candidate drug molecule on thousands of proteins, including TFs in live cells, and build up a dataset of how compounds impact protein activity in different cellular contexts, such as different cancer types, all in their native cellular environments.

This is different than what other companies that have gone after transcription factors have done, as they have used artificial systems that divorce these proteins from their normal environment, Federation explained.

"Previous technologies attempted to reconstitute artificial environments to study transcription factors, but transcription factor proteins only fold and function properly inside the cell, which is a major reason why traditional drug development approaches have failed for these targets," he said.

"We realized that by deploying cutting-edge technologies, we could overcome this problem and study these transcription factors in live, unmodified human cells."

MARMOT also uses new AI models that are trained on Talus Bio’s dataset for transcription factor activity; the company spent the last year amassing what it believes to be the largest proprietary dataset of the effects of drug-like molecules on transcription factors. The company has now leveraged this data and combined it with foundational models built by the AI community to train its AI system to accelerate TF drug discovery.

That means the model now can predict the impact of any drug-like molecule on any TF in the cell, so it can identify optimal candidates from billions of compounds much faster, replacing high-throughput screens, which have historically have been time consuming, and which have failed for TFs.

"In the past 30 years, drug discovery has relied on high throughput screening where scientists screen millions of molecules, chosen at random, to find starting points for therapeutic development. That’s not efficient. Instead, we can take advantage of what we’re learning with the MARMOT platform and have our AI models suggest which molecules are the best starting points for the most important transcription factors contributing to disease," said Federation. 

"In terms of ROI, this will provide a massive boost in the value-per-screen at the bench, decreasing the number of compounds that need to be tested to deliver a successful drug, ultimately leading to massive savings in time and cost as well. Ultimately, we want to get the best therapy possible to patients as soon as possible, whether from our internal pipeline or working closely in collaboration with larger pharma partners."

The new funding will be used to advance Talus Bio’s therapeutic programs, which include Brachyury-driven cancers like Chordoma, a rare spinal cord cancer, and an undruggable TF implicated in prostate cancer. The funds will also be to accelerate the company's work on its MARMOT platform and expand the capabilities of the AI models that are trained from its data.

The money will also be used to expand Talus' data science team to accelerate the training and deployment of its machine learning models for drugging TFs. These algorithms will inform continued expansion of its internal chemical library, rapid optimization for itsinternal programs, and new compound discovery for external pharma partners.

"By increasing our focus on our AI/ML team we believe we’ll accelerate drug discovery, and also accelerate the path to clinic," said Federation. 

Along with the funding, it was also revealed that J. Seth Strattan, PhD, General Partner at Two Bear Capital, be joining the board of directors at Talus Bio. 

"Dr. Strattan is a scientist and entrepreneur who brings deep expertise in transcription factor biology and data science. Prior to his role at Two Bear, he led engineering and data science teams in several large consortia focused on understanding the fundamental biology of genome regulation," Federation said.

Ultimately, the goal at Talus is to enable the discovery of a drug for every transcription factor that needs one.

"We’ve seen this story before – Genentech and others unleashed new technology to make the cell surface druggable with antibodies, leading to over 100 drugs approved in this class today. Kinases were unlocked in 2001 and now have nearly 100 drugs approved. We see a future where our technology can unlock disease-causing transcription factors at a similar scale," Federation said.

(Image source: talus.bio)