Mutations in the KRAS protein are among the most common drivers of many cancers, including pancreatic, colorectal, and non-small cell lung cancers. KRAS acts as a molecular switch, and when it’s stuck in the “on” position, it drives uncontrolled cell growth. SOS1 is a key activator of KRAS, making it a compelling upstream target for indirect KRAS inhibition. SOS1 inhibitors may also enhance the effectiveness of direct KRAS inhibitors and other MAPK pathway agents, potentially extending the therapeutic window and delaying resistance.
When we began our SOS1 inhibitor program, there were already reported molecules from two companies targeting the same mechanism. At first glance, it looked like there was only one way to design a potent SOS1 inhibitor — all the other existing molecules were making the same three key interactions in the binding pocket. In fact, when we reviewed the leading candidates in the industry over the course of our program, many of them were chemically highly similar structurally, often differing by a single nitrogen atom change. We wanted to design a truly novel, differentiated molecule, which meant exploring regions of the binding pocket that had been overlooked, dismissed, or deemed too difficult to target.
Necessity Sparks Invention
At the onset, we knew the rules of medicinal chemistry had long discouraged targeting solvent-exposed residues at the edges of binding sites. Every new chemist reads the classic paper, A Medicinal Chemist’s Guide to Molecular Interactions, which essentially states: don’t try to make meaningful interactions with amino acids that are fully exposed to water — it doesn’t work. These are the “coastlines” of the protein, surrounded by a sea of water, and attempting to form strong interactions there is considered a waste of time due to factors like electrostatic screening and desolvation penalties.
However, when we closely examined the SOS1 binding site, we saw three amino acids that had not been touched by any of the existing inhibitors. These were the “leftover” interactions, sitting out on the coastline. Conventional wisdom told us to ignore them. But conventional wisdom wasn’t going to get us where we needed to go.
Leveraging Schrödinger’s Platform and FEP+
This is where Schrödinger’s computational platform, particularly Free Energy Perturbation (FEP+), gave us a competitive advantage. We believed that if there were a way to make these edge-case interactions work, FEP+ would help us find it. Unlike traditional methods that rely on guesswork or brute-force synthesis — or AI/ML models that are fundamentally limited by their reliance on historical data — FEP+ allowed us to accurately simulate and predict which molecules had the potential to form high-affinity interactions with the untouched amino acids near the edge of the binding site, even when those interactions had never been seen before.
FEP+ identified a molecule with a nitrogen-containing cluster of atoms that carried a positive charge, enabling the molecule to form a salt bridge, or strong ionic bond, with two negatively charged amino acids on the protein surface that were previously considered untargetable. This molecular design resulted in a ~750-fold improvement in binding affinity compared to the same molecule made entirely of carbon atoms. We confirmed the protein–ligand interaction using X-ray crystallography and showed that binding strength was closely linked to the presence of a strong ionic bond between the molecule and the protein, exactly as predicted by FEP+. This unconventional approach allowed us to break into new chemical space, ultimately resulting in the discovery of SGR-4174, our IND-ready SOS1 inhibitor.
Reactions From the Scientific Community
Within a few months of its publication, our manuscript has already been cited by researchers around the globe, including those from China, Europe, USA, and the UK, demonstrating its recognition and acceptance by the broader, international scientific community. We were especially encouraged to see a review of our work by Antonia F. Stepan, Section Head, Medicinal Chemistry at Roche and Nicholas Meanwell, distinguished professor and 40-year Bristol-Myers Squibb retired Senior Vice President of Chemistry:
“I found this study interesting because it demonstrates that engaging charged side chains on the solvent exposed surface of a protein with a complementary charged element in a ligand using a salt-bridge interaction can lead to enhanced affinity, refuting the generally accepted expectations for this kind of molecular edit,” Meanwell said of our work. “The authors provide theoretical insights into the phenomenon and results have broader implications in drug design, suggesting that engaging surface-exposed charged side chains in salt bridge interactions can confer benefit, which should encourage others to explore this approach to enhancing drug-ligand affinity.”
I’ve enjoyed presenting the research at the Applied Pharmaceutical Chemistry meeting in Waltham, MA, the Free Energy Workshop in Cambridge, MA and the New York Structural Biology Discussion Group in New York, NY. It’s been inspiring for me to see the enthusiasm and interest with which our results have been received. The different audiences at each of those meetings — medicinal chemists, computational chemists, and structural biologists, respectively — have each been receptive to our findings, interpreting them through the prism of their own training and experience.
Broader Implications Beyond SOS1
This isn’t just a one-off phenomenon. There is an emerging precedent across drug target classes where carefully placed interactions with solvent-exposed residues led to dramatic potency gains. For example, in a specific KRAS mutant known as KRAS G12D, a similar surface interaction led to a 1,000-fold gain. A similar phenomenon appeared with HPK1 and PRMT5, two other oncology targets. In these three cases, we confirmed retrospectively that FEP+ would have accurately predicted the gains in potency.
We haven’t overturned a basic law of medicinal chemistry, and it’s certainly true that in most cases one should not expect to gain potency with solvent-exposed saltbridges. But equally, we have shown that the “rules” aren’t absolute. While most interactions with solvent-exposed residues won’t work, the right one can make all the difference. Tools like FEP+ can help us find those rare but powerful opportunities.
As drug discovery moves into increasingly challenging territory, teams will be forced to think creatively. SGR-4174 is proof that bold, inventive design powered by computation can unlock opportunities others overlook. A “predict-first” approach enables teams to uncover novel interactions, access untapped regions of chemical space, and confidently pursue targets once deemed undruggable or too crowded to enter. It’s not just about finding molecules — it’s about changing the way we discover them.
