A computer scans millions of proteins that parasites use to hide from the immune system and picks out the ones most likely to work as medicines for autoimmune diseases.

Ditto Biosciences has built a proprietary AI platform that systematically mines the genomes of parasites—including viruses, ticks, and worms—to identify proteins that have evolved over millions of years to modulate the human immune system. The platform ingests genomic and proteomic data from diverse parasite species, applies sequence-function modeling, protein structure prediction (likely leveraging transformer-based architectures such as ESM or AlphaFold-derived models), and binding affinity estimation to rank and prioritize candidate proteins. Within seven months of operation, the platform analyzed over one million proteins and identified thousands predicted to interact with validated human immune targets, many exhibiting drug-like binding affinities. This approach fundamentally differs from traditional drug discovery by starting not from human biology or synthetic chemistry, but from the evolutionary solutions that parasites have already optimized over millennia. The result is a vast, novel chemical space of protein therapeutics that conventional pipelines would never explore.

It's like hiring a million-year-old parasite as your drug design consultant—it already figured out how to calm down the immune system, and the AI just translates its notes.

An AI acts like a protein personal trainer, tweaking and improving parasite proteins until they're strong enough, stable enough, and safe enough to become real medicines.

Once Ditto's discovery platform identifies promising immunomodulatory proteins from parasite genomes, the next critical step is engineering those proteins into viable drug candidates. Ditto employs ML-guided directed evolution and generative protein design to optimize leads for key pharmaceutical properties: binding affinity to human immune targets, thermostability, solubility, half-life, and manufacturability. The platform likely uses variational autoencoders (VAEs), diffusion models, or protein language model fine-tuning to propose sequence variants that improve multiple properties simultaneously—a multi-objective optimization problem that is intractable through traditional wet-lab screening alone. By coupling computational predictions with rapid experimental validation in a design–build–test–learn cycle, Ditto can explore orders of magnitude more sequence space than conventional directed evolution, dramatically accelerating the path from hit to lead to preclinical candidate. This ML-guided approach reduces the time and cost of protein engineering while increasing the probability of identifying candidates with truly drug-like profiles.

It's like using AI to speed-run evolution—instead of waiting a million years for nature to perfect a protein, you get the optimized version by Friday.

An AI cross-references a library of real human immune tissue samples against candidate drug proteins to predict whether a patient's body would reject the medicine before it's ever tested in people.

One of the greatest challenges in developing novel protein therapeutics is immunogenicity—the risk that the human immune system will recognize the drug as foreign and mount an immune response against it, reducing efficacy or causing adverse reactions. Ditto Biosciences addresses this by building a proprietary tissue biobank that captures real-world immune memory profiles from diverse human donors. The biobank data is integrated with ML models that predict T-cell and B-cell epitopes, MHC binding, and cross-reactivity patterns for each candidate protein. By mapping the immune landscape against their parasite-derived candidates, Ditto can computationally flag high-risk sequences and use protein engineering to redesign or de-immunize candidates before they ever enter preclinical testing. This data-driven immunogenicity prediction loop is a critical competitive advantage: it allows Ditto to de-risk candidates earlier, reduce costly late-stage failures, and design proteins that are not only effective but also well-tolerated by the human immune system. The approach is especially powerful given that parasite-derived proteins are inherently foreign to the human body, making immunogenicity prediction not just valuable but essential.

It's like checking if your body's bouncers will recognize and kick out the new drug before you even send it to the club—saving everyone a very expensive night out.