5′— A U G • G C C • U A C —3′
RNA code Designing RNA medicines
mRNA and RNA therapeutics designed end to end — untranslated regions, codon choice, secondary structure, and stability — optimized for expression, durability, and delivery.
AUG ACG GUC AAA CGC UAC GGC UCC AAG CGU AUG GCC CUG GAC AUC GGC
CODE
Center for AI-Designed Genetic and Cellular Encoding
A national center for AI-designed medicine
Building foundation AI for programmable RNA, gene, and cell therapies.
The future of gene and cell therapy is written in code — RNA code, genetic code, immune-recognition code, and cellular reprogramming code.
RNA Genetic Immune-recognition Cellular reprogramming
01The premise
Gene and cell therapy is the proven frontier of medicine — mRNA reached billions, siRNA is an approved drug class, CAR-T delivers durable cures. The modality works. What's missing is the design substrate: a general way to design across the vast space of sequences, and a general way to predict who a therapy will work in. These are foundation-model problems — not solved one program at a time. CODE builds that substrate as a national center for the code of life.
02Four codes of life
One center, learning every layer of biological code.
… TATA • CCAAT • enhancer …
Genetic code Engineering DNA as software
Regulatory elements, promoters, and gene circuits — plus the guides and edits that write durable, controllable function directly into the genome.
TCR ⟷ peptide–MHC
Immune-recognition code Reading the immune system
How T cells and antibodies recognize the body — antigens, epitopes, and TCR–pMHC binding — modeled and designed for precision immunotherapies and vaccines.
factors → cell state Δ
Cellular reprogramming code Rewriting cell identity
The transcription-factor programs that switch a cell's state — turning one cell type into another and tuning cells to repair, replace, and heal.
03The platform
Two missing substrates — and the loop that builds them.
A design layer over the whole sequence space, a response layer that predicts who a therapy works in, and a closed loop that makes both better every cycle.
sequence → representation → design
Design layer Genomic foundation models
Language models of nucleotide sequence that read and generate DNA and RNA across the full design space — building on the DNABERT and GenomeOcean lineages.
multi-omics · imaging · transcriptomics
Response layer Multimodal virtual-cell models
An in-silico assay and in-silico patient that predict cellular response before the experiment — and group patients before the trial (Cell-JEPA lineage).
design → run → read → repeat
The closed loop A self-driving lab
An active-learning agent designs, runs autonomous wet-lab experiments, reads them out, and chooses the next — generating sequence-to-function-to-response data no dataset holds, and improving every model each cycle.
04What we design
One substrate, every modality.
Silencing, by design
Sample-efficient design of small interfering RNA, with knockdown and selectivity built in.
Coding and control, together
Joint generative design of coding sequence and UTRs for higher expression and durability — not stitched-together pipelines.
Where the full stack meets
Engineered cell therapies designed end to end and matched to the patients most likely to respond.
The right patient, not just the molecule
In-silico cohorts so every therapy carries a hypothesis about who it helps — most failures are the wrong patient, not the wrong design.
05Join us
We're assembling the team writing the next code of medicine.
Researchers, engineers, and partners across AI, RNA, genome, and cell engineering — at Northwestern and nationwide.
Get in touch