Mankind and Denovo Partner as AI Drug Development Enters a New Stage of Human-Machine Review

The most expensive failures in new drug development often do not occur in the final mile of clinical development, but much earlier: a seemingly promising molecule, after synthesis, testing, and repeated modification, only then reveals insufficient activity, toxicity concerns, or poor developability. The new collaboration between Mankind Pharma and Denovo Sciences is aimed precisely at this long and uncertain early drug discovery process.

According to reports from ETPharma and The Economic Times, Indian pharmaceutical company Mankind Pharma has signed a collaboration with Denovo Sciences to launch an AI-driven drug discovery program. The goal described publicly by the two parties is to shorten early R&D timelines, improve the quality of lead compounds, and screen out candidate molecules with a greater likelihood of advancement at an earlier stage.

The division of roles in this collaboration is quite clear: Mankind will provide R&D infrastructure, along with experimental and clinical validation capabilities; Denovo will introduce its customized AI platform to generate and evaluate molecular candidates. In other words, AI is not being used here simply to produce a batch of attractive chemical structures, but is being placed inside an R&D loop that must be examined, revised, and revalidated by scientists.

The reports noted that the two parties will adopt a “human-in-the-loop” model, in which AI proposes and scores molecular designs, and researchers then make judgments based on biological plausibility, synthesizability, and subsequent validation results. This design reflects the more pragmatic side of AI drug development today: models can expand the search space, but whether something can become a drug still has to be tested against experimental data, disease mechanisms, and clinical translation.

Public information remains limited for now. The reports did not specify the disease areas targeted by the collaboration, the types of targets, the sources of training data, the methods of model validation, or whether any specific candidate molecules have entered experimental testing. These gaps matter because the real threshold for AI drug discovery often lies not in whether molecules can be generated, but in whether the generated molecules can be synthesized, whether they can reproduce the expected effects in cell or animal models, and whether they can hold up in terms of safety and pharmacokinetics.

Background Context

In recent years, collaborations between pharmaceutical companies and AI drug platforms have become increasingly common, but industry evaluation standards have also gradually shifted from “whether AI is being used” to “where in the R&D process AI produces verifiable value.” If the Mankind and Denovo program can disclose more specific experimental progress, such as improved hit rates, faster lead compound optimization, or comparisons with traditional screening processes, it will be better able to show whether this type of collaboration has reproducible R&D efficiency.

For regulation and clinical development, AI itself will not give candidate drugs a shortcut. If molecules advance to the preclinical or clinical stage in the future, they will still need to submit evidence of quality, safety, and efficacy according to existing regulations. This collaboration therefore looks more like an experiment in transforming the front-end R&D process: algorithms can accelerate proposals, but what ultimately remains must be candidate drugs that can withstand biological and medical validation.