Bringing sequence design out of the computer

Every day, it seems, there’s exciting news in biotech about new AI models, new startups, and new tools for simulating and designing biological sequences and structures. But even as it gets easier and easier to simulate biology, the work of actually building molecules, developing and manufacturing real medicines remains incredibly challenging, in ways no model can yet predict.

At the end of the day, a therapeutic molecule has to exist in real life, but there is still a gap between in silico predictions and in vivo or even in vitro performance that biotech companies trying to use these models encounter every day. Computationally promising digital sequences and molecules can’t always be manufactured and delivered successfully. 

With RNA, scientists can model sophisticated sequences on the computer, but at the bench, these sequences suffer from degradation alongside variability in synthesis, purity, and integrity. The challenge gets even harder when we try to understand how these molecules function inside complex biological systems, where stability, expression, and immune response are key readouts but remain difficult to model. We need new approaches to bridge the gap between artificial intelligence and biological reality. 

We need better therapeutics faster, and while AI promises to accelerate R&D iteration, models alone aren't enough—they need to be guided by real-world experiments generating high quality data. Truly powerful models can only emerge from bringing our designs out of the computer quicker. Real progress in our ability to design effective RNA medicines will come from tight coupling between computational design and experimental validation – what we call "ex silico" development.

Bringing therapeutics out of the computer

mRNA has massive potential as a platform for therapeutics thanks to its promise of programmability. Who wouldn’t want to use the body’s own machinery to create the perfect protein to solve the problem it faces? But underlying that potential is a complex interplay: nucleotide sequence, chemical modifications, and structure all work in concert to power the transcription of genetic material into amino acids. Often, it isn’t what we already know about these elements that causes issues: it’s what we don’t know about them that derails progress.

The current reality is that there are no purely computational shortcuts in biology. Every sequence that gets designed must be manufactured, tested, and validated in the real world. 

Only by generating data from actual molecules can we get a handle on how these molecules are likely to perform. Ex silico development is the key to understanding these parameters and providing a compass to rapidly iterate and generate smarter shots on goal.

A race against time

Solving the challenges of drug development requires iterating through the design-build-test cycle at high speed without sacrificing quality. The faster you can actually manufacture a molecule and test its properties in the real world, the more chances you have of finding a molecule that doesn’t just look good in the computer. Reducing this cycle time is crucial to maximizing your chances of success. 

With ex silico development, Terrain puts more time on the clock.

If you need better RNA, not just a better sequence prediction, please reach out.