AI-Designed Drugs Face a Clinical Test: Takeda TYK2 Inhibitor Wins in Psoriasis Comparator Trial

For AI drug development, the truly difficult part is often not generating an elegant molecule, but proving how much of a difference it makes in humans, alongside standard treatment. Takeda Pharmaceutical’s oral TYK2 inhibitor zasocitinib has now crossed that threshold, one closer to clinical reality, in a trial for moderate-to-severe plaque psoriasis: according to Investor's Business Daily, the drug, initially designed on the Nimbus platform and later acquired by Takeda, achieved a markedly higher rate of complete skin clearance after 16 weeks than Bristol Myers Squibb’s marketed drug Sotyktu.

The core of this comparison is not whether a placebo group lost to a new drug, but a face-to-face competition between two oral drugs in the TYK2 pathway. In data Takeda released in March, the company referred to zasocitinib as TAK-279, a “next-generation, highly selective” oral TYK2 inhibitor being developed for moderate-to-severe plaque psoriasis; the same announcement also mentioned that the company was conducting a head-to-head study against deucravacitinib. Sotyktu is the brand name for deucravacitinib.

The ClinicalTrials.gov registration NCT06973291 adds the trial outline: the study population consists of adult patients with plaque psoriasis, and the study compares zasocitinib (TAK-279) with deucravacitinib. This makes the clinical significance of the latest news clearer. For patients, the appeal of oral immune-modulating drugs lies in avoiding the inconvenience of injections; for physicians and payers, the question is whether a new drug can deliver an efficacy gap large enough to change prescribing order beyond existing options.

TYK2 is an intracellular signaling enzyme involved in multiple inflammatory cytokine pathways. Inhibiting this pathway can reduce abnormal immune activation in psoriasis, but selectivity has always been central to drug design because related JAK family signaling is also involved in broad immune functions. AI’s specific role here is not to directly diagnose disease or replace clinical judgment, but to help search for and optimize candidate molecules that can act precisely on the target protein during the early chemical design stage; ultimately, human trials still have to answer questions of efficacy and safety.

Takeda’s LATITUDE PsO Phase 3 trial, released in March, had already shown that zasocitinib met skin clearance-related endpoints at week 16 compared with placebo and apremilast, providing background for the later comparison with deucravacitinib. If the reported head-to-head results hold up in the complete data, including details such as different doses, discontinuation rates, adverse events, and long-term response, zasocitinib would be more than an “AI design success story”; it would be a candidate drug that could potentially reshape the competitive landscape for oral psoriasis treatment.

However, the public summary still leaves several important blanks. The report said the complete skin clearance rate was higher, but the full data, statistical design, baseline differences among patients, and overall safety profile still need confirmation through a formal conference presentation or journal article. Psoriasis drugs are also not judged only by clearance rates; durability of efficacy, risks of infection or laboratory abnormalities, and whether they have a clear position compared with biologics also matter.

This is also where the AI drug development narrative is shifting. The market may pay for “AI design” for a time, but regulators and clinical practice still look at traditional questions: whether the trial is rigorous enough, whether benefits outweigh risks, and whether the data can support approval for a specific indication. If zasocitinib continues to advance, it will become a concrete case for testing whether AI-assisted molecular design can be translated into a drug usable by patients, rather than remaining only an algorithm story.