Designing a new class of precision molecules
We design oligonucleotide binders against a chosen site on a protein, computed from its own sequence. One molecule class can block a protein interaction, reach a target inside the cell, or report whether the protein still works.
Designed, not screened
A screen only returns what happens to bind, which is why flat interfaces and disordered regions come back empty. We start from the protein, pick the site, and compute the binder. If a design fails, we go at a different site. Days, not another campaign.
Therapeutic capabilities
A designed binder stopped a disease-driving protein complex from assembling, since replicated independently at a large London hospital. A second was taken up into cells. Neither has been into an animal.
Human validation
The same method produced a published biomarker. Our TDP-43 assay reads whether the protein still works, not how much is there: 1,080 serum samples, AUC 0.79 separating ALS from controls. Biosensors, August 2026, with co-authors from Mass General Brigham, Johns Hopkins, the NIH, Barrow and ICGEB
One platform. Multiple functions
Block a protein–protein interaction. Reach a target inside the cell. Measure how much of a protein is there, what state it’s in, and whether it still works. One design method, pointed in different directions. The molecule that changes what a protein does and the molecule that reports it come out of the same run.
PLATFORM SCIENCE
Start with the protein site. Design the molecule. Test the function.
- Design-first > Proteins act through specific surfaces and interactions. We use the sequence and a chosen site to design a binder for that region.
- Not screened > Conventional discovery builds a library and searches for what binds. We reverse it: choose the site, design to it, test whether it does the job.
- Built for therapeutic function > The same logic points at protein interfaces or at targets inside the cell. Both demonstrated in vitro.
- Proven in human blood > A TDP-43 functional assay across 1,080 human samples, published in Biosensors in 2026. Sequence-designed molecules can read functionally relevant biology in blood.
The following comparison illustrates how our innovative approach stands apart from existing RNA technologies:
Nemdx synthetic RNA
- Protein, DNA, RNA targets
- Engineered from amino-acids
- Multiple Tx Modalities
- Measure function & quantity
- Synthetic scalable production
miRNA / siRNA
- DNA, RNA targets
- Controls protein production
- Regulates gene expression
- Cellular machinery dependent
- Cellular factors needed
Impact
- Optimized binding to proteins
- High sensitivity/specificity
- Broad clinical applications
- Reliable reproducibility
- Streamlined development
Molecular Technology
Small oligonucleotide binders, computed from the target’s own sequence against a chosen site.
-
Sequence-designed oligonucleotides
We map each amino acid across the chosen region to its cognate nucleotide and write the chain for that site. The 2′-O-methyl backbone is what lets it survive in blood and inside cells. -
Function is part of the design brief
We decide what the molecule must do before we design it. Screens optimise for binding, and binding isn’t doing: antibody campaigns against hard targets have produced molecules that engaged the protein and changed nothing.
-
A common platform for drug and biomarker development
The molecule that changes what a protein does and the molecule that reports whether it worked come from the same design run, against the same site. Alzheimer’s spent tens of billions on a drug and a biomarker measuring different things. Shown on TDP-43. Not yet built for a new target.
Design Process
From a named target to a functional answer in about six weeks.
-
Choose the target and functional site
You name the protein, the region that matters and the effect you want. Because the site is chosen rather than returned by a screen, it can be somewhere a screen never reaches. - Design from sequence About a day > Each amino acid maps to its cognate nucleotide. No library, no selection. Off-target risk is checked against the proteome while the molecule is still a file.
- Synthesize the defined molecule Two to three weeks > One molecule, not a pool. Nothing to triage..
- Test biological function The endpoint is the biological effect, not a binding curve.
- Advance with the partner If it works, the data sets the next step. If it doesn't, we design against a different site on the same protein — another design run, not another campaign. That's the part that changes the economics of a hard target.If it works, the data sets the next step. If it doesn’t, we design against a different site. Another design run, not another campaign.
One platform. Multiple functions.
Nemdx's Recent
News & Other Resources
NEMDX
Contact Us
NEMDX
Locations
- Headquarters: Boston, MA, USA
- R&D Lab: Discovery Park, Sandwich, UK
- Commercial Lab: Framingham, MA, USA