Schrödinger’s Akinsanya On Low-Risk Pipeline Growth And Platform Evolution

Since its founding in 1990, New York-based Schrödinger has functioned as a platform company, identifying, validating, optimizing clinical candidates and doing preclinical development for other companies. But the company has in the past few years been employing its platform to build a pipeline of first-in-class or best-in-class therapies.

Since Karen Akinsanya took on the role of president of R&D of its therapeutics division two years ago (after four years as chief biomedical scientist at the company), her team has taken three programs into the clinic and have another five disclosed programs in preclinical or discovery stages.

“Our vision is to span both first-in-class, best-in-class and potentially best-in-disease mechanisms, to demonstrate that we can come up with excellent molecules, because we are leveraging the Schrödinger platform, and to take a handful of them into the clinic to generate value for patients, healthcare stakeholder groups, and the company,” Akinsanya told In Vivo.

Schrödinger is not the first platform company to try its hand at drug development and results have been mixed. BenevolentAI, for example, has been effective in drug discovery but provided disappointing results for its own pipeline and has faced leadership shakeups as a result.

Low-Risk Approach To Wholly-Owned Pipeline

However, Schrödinger is taking a risk-averse approach – at least initially—to building its pipeline. Indeed, when it began developing its wholly-owned pipeline six years ago, “we wanted to initiate our pipeline with targets where there was some evidence in humans that the mechanism worked,” Akinsanya explained.

They chose precedented targets and pathways that had existing treatments where there was “a gap in clinical efficacy or duration of response.” Where both a validated target and a suboptimal commercially-available treatment existed, Schrödinger developed potential combination agents, specifically for oncology.

Our overarching view is that if you can add another small molecule agent for the majority of patients who are relapsing or resistant to improve duration of response or improve the number of patients in remission, that’s an exciting goal.

Karen Akinsanya

“Our overarching view is that if you can add another small molecule agent for the majority of patients who are relapsing or resistant to improve duration of response or improve the number of patients in remission, that’s an exciting goal,” she said.

Today, Schrödinger’s lead asset is its small molecule mucosa-associated lymphoid lymphoma translocation protein 1 (MALT1) inhibitor, SGR-1505. Johnson and Johnson Innovation has yielded Phase I proof of concept that MALT1 inhibition has efficacy in non-Hodgkin’s lymphoma as monotherapy and increases efficacy when given with a BTK inhibitor in chronic lymphocytic leukemia patients.

In healthy subjects, SGR-1505 was generally well tolerated with no dose-limiting toxicities or serious adverse events, and it had a favorable pharmacokinetic profile and evidence of target engagement.

The two other clinical-stage assets are a cell division cycle 7-related protein kinase (CDC7) inhibitor for hematologic malignancies and solid tumors (SGR-2921) and a Wee1/membrane associated tyrosine/threonine 1 (Myt1) inhibitor (SGR-3515), for solid tumors.

Those two agents are in ongoing Phase I studies and the company expects to present initial Phase I data for all three molecules in 2025.

Starting To Think Bigger

Now that Akinsanya and her team have moved the three programs into the clinic as add-ons to standard of care regimens, they are starting to think bigger, and riskier.

“If you ask me about our vision today, I think we feel that having built the clinical team, we can dig deeper into questions of unmet need, therapeutic opportunity beyond standard of care, or even targets where there is no standard of care,” she said.

If you ask me about our vision today, I think we feel that having built the clinical team, we can dig deeper into questions of unmet need, therapeutic opportunity beyond standard of care, or even targets where there is no standard of care.

Karen Akinsanya

The company is exploring whether SGR-1505 may have efficacy in certain dermatologic as well as hematologic indications, but it is still sticking to an approach that favors precedence over novel discovery. There is genetic evidence that reducing Nuclear factor (NF) kappa B signaling can provide clinical efficacy for certain dermatologic conditions, and MALT1 inhibition, “does a really good job at dampening NFkB signaling,” she said.

“I think we feel much more confident about taking a little bit more risk –not a lot of risk. You’re not going to see us pursuing targets out of nature,” she emphasized. “We think it’s just too risky when we don’t have orthogonal validation, either from the clinic or human genetics, that shows a gene-dose response curve.”

This approach still leaves room for novelty, however. For example, Akinsanya is open to developing first-in-class small molecules where there is already a biologic.

And one of the clinical assets Akinsanya is excited about, its Wee1/Myt1 inhibitor, has potential to act via a mechanism of action that the biopharma industry has been looking to replicate since the success of PARP inhibitors, namely synthetic lethality.

“If we identified a synthetic lethal relationship, like Wee1/Myt1, we would be open to taking a first-in-class play,” Akinsanya said.

Partnerships Key

Signaling confidence in the team she has built up over the past six years, Akinsanya said Schrödinger would be willing to develop an efficacious agent alone if it saw first-in-class potential where there was no standard of care. Doing so would require costly investments into clinical capabilities, such as designing endpoints, building patient cohorts and establishing biomarkers.

More likely, however, is that Schrödinger stays focused on “building strong capabilities, structurally enabling targets that we think will be the cream of the crop over a longer period in our pipeline and, for those targets where there is a clear path to establish excellent behavior in Phase I and get proof of concept data, the company retains the option to pursue initial development on our own,” Akinsanya said.

Schrödinger will likely move assets along in the clinic to a point, and then partner with companies “that are much better positioned to expand into other indications or to lead the regulatory strategy.”

That said, the company is building a development team and has hired a chief medical officer, Margaret Dugan, formerly CMO of oncology company Dracen Pharmaceuticals and SVP at Novartis.

The company is positioning itself for “a value inflection opportunity with one of our programs that emerges from the pipeline,” Akinsanya said.

Schrödinger’s Platform In The Age Of AI

The engine of Schrödinger’s success, both in terms of serving others and in building its small pipeline to date, is its computational physics-based platform. The company is not thought of as an AI-driven drug discovery company, in the same vein as Recursion Pharmaceuticals, BenevolentAI, or InSilico, for example, and indeed its platform predates the widespread use of AI.

Akinsanya is confident Schrödinger will maintain its position as a trusted computational drug discovery company and software provider for the industry. In a recent corporate presentation, the company reported serving over 1,785 software customers, including 20 top biopharma companies, and drug discovery revenues rose from $5.8m in 2023 to a forecasted $11.9m in 2024.

Akinsanya said the company has built “a pretty big moat” around its proven physics-based approach and ongoing investments in the platform and new breakthroughs make “the overall suite of solutions we offer pretty difficult to replicate.”

Augmenting With Machine Learning

When American chemist Richard Friesner co-founded the company in 1990, predicting molecular properties was slow and arduous but today’s computing power – Schrödinger had been a heavy user of NVIDIA gaming chips for a long time and now employs some of NVIDIA’s most advanced tools –means the process can be done rapidly.

Akinsanya said Schrödinger’s technology evolved significantly in 2015, when programmers began coding approximations of free binding energy into software. Schrödinger’s own programmers coded for physics-based interactions, allowing the platform to calculate the affinity of small molecules and proteins on the atomic level – using a highly accurate version of what experts in this space call a “Force Field.”

“Being able to perform physics-based calculations that are consistently accurate and very scalable can be seen as a computational assay,” said Akinsanya, suggesting in silico assays are a time and cost-saver in the early stages of drug discovery and development.

The company has also augmented its physics-based approach with machine learning tools. While physics-based predictions “are highly accurate but a little bit slow and computationally expensive,” so adding ML has meant the company “can now enumerate vast amounts of chemical space with these free-binding calculations.”

For the company’s MALT1 program, for example, the team screened physics-based interactions with 12 billion molecules.

Unlike some other companies, Schrödinger uses ML in a “very target-specific way.”

“Mother Nature was very thrifty in some ways, reusing folds and domains across proteins but also created this exquisite diversity spanning around 20,000 unique proteins in our bodies,” Akinsanya noted. “There can be overlap with other proteins, but the physics-based models that we create and scale with ML for a target are specific to that target.”

The Era Of The Proteome

Importantly, the growing knowledge base surrounding protein structures and behavior thanks to tools like AlphaFold is good news for Schrödinger. It is thought that only 10-15% of protein structures are solved experimentally, but generative AI is helping improve that situation. Alphafold, whose inventors were awarded with a 2024 Nobel Prize in Chemistry, can predict protein folding with high accuracy, but the macro-level structures revealed by AlphaFold are only a starting point for the biopharma industry.

Once we have an accurate picture of a protein structure, from AlphaFold or other sources, using our physics approach, we can refine those structures to the point where you can use them for atomistic simulations.

Karen Akinsanya

“Once we have an accurate picture of a protein structure, from AlphaFold or other sources, using our physics approach, we can refine those structures to the point where you can use them for atomistic simulations,” she said.

While the 1990s and early 2000s were the decades of the genome, Akinsanya asserted we are in the era of the proteome, “and that’s good for Schrödinger,” said the CEO.

“If you think about that small portion of the human proteome that we had access to, now there is a rapidly accelerating substrate for our platform, thanks to our refined structural models,” she said. “We find that particularly exciting.”

While tools like AlphaFold3 can predict protein folding, published literature suggest it cannot meet the sophisticated needs of drug discovery.

“In drug discovery, you’re not figuring out how known molecules like ATP binds, which is what AlphaFold3 replicated in the recent publication. You’re figuring out if there is a drug molecule that binds and has all the right properties, solubility, permeability, brain penetration,” she emphasized.

The next stage of evolution for Schrödinger’s own computational assays is calculating free binding energy without a reference molecule. That is a task Akinsanya said, “we’ve made huge progress on.”

Straddling Two Worlds

In a way, when it comes to developing its pipeline and its platform, Schrödinger straddles two worlds: while it is making relatively safe bets in its pipeline strategy, when it comes to enhancing its platform, “we don’t do incremental, we go for very big, grand challenges.”

We don’t do incremental, we go for very big, grand challenges.

Karen Akinsanya

With a strong industry reputation and strong track record, that platform will likely continue to serve other companies, even as a flood of AI-driven drug discovery newbies emerge. Whether the same platform can translate to a successful pipeline, and the company can carry out the clinical development that it has not to date, remains to be seen.