The company behind TikTok’s addictive recommendation system is now applying its artificial intelligence expertise to one of the toughest challenges in medicine: creating drugs for targets long dismissed as “undruggable” by traditional pharmaceutical approaches.
ByteDance, the Beijing-based tech giant that owns TikTok, has established Anew Labs (also operating as Anew Therapeutics), a dedicated drug discovery unit focused on using generative AI to develop small-molecule therapies. On May 2, 2026, reports highlighted the unit’s latest milestone: presenting preclinical data on an AI-designed therapy at a major immunology conference.
Targeting the “Undruggable” with AI
In mid-April 2026, Anew Labs presented at the American Association of Immunologists’ annual meeting in Boston. They showcased a generative AI-designed small molecule that inhibits IL-17, a cytokine central to autoimmune and inflammatory conditions such as psoriasis, rheumatoid arthritis, and ankylosing spondylitis.
Existing IL-17 therapies, including blockbuster injectable antibodies like secukinumab (Novartis) and ixekizumab (Eli Lilly), generate billions in annual revenue. However, these treatments require injections and can be expensive. Anew’s candidate is a small molecule designed as an oral pill capable of blocking multiple IL-17 forms (pan-spectrum inhibition).
The target is particularly difficult because IL-17 interacts with its receptor via a broad, flat protein-protein interface. Such surfaces offer little for conventional small molecules to grip, which is why the industry has historically labeled many of these interactions “undruggable.” Anew Labs claims its AI found an effective way to disrupt this interaction.
The Power of AnewOmni and Cross-Scale Design
Central to the effort is AnewOmni, a generative AI framework trained on more than five million biomolecular complexes. Unlike many models limited to one molecular scale, AnewOmni reportedly works across small molecules, peptides, and nanobodies. It uses programmable graph prompts to incorporate chemical, geometric, and topological constraints during molecule generation.
Preprint results demonstrated the model’s ability to design functional molecules against challenging targets like KRAS G12D (a notorious oncology driver) and PCSK9 (involved in cholesterol regulation), with laboratory validation success rates ranging from 23% to 75%.
Anew Labs operates from Shanghai, Singapore, and San Jose, California. Its scientific advisory board includes veterans from Innovent Biologics, Amgen, and Takeda, bringing deep expertise in biologics and immunology. The unit’s broader pipeline includes candidates against IL4R and other undisclosed targets, with ambitions to accelerate discovery from target identification through clinical development.
From Viral Videos to Molecular Prediction
There is a conceptual parallel between TikTok’s algorithm and Anew’s drug design tools. Both process enormous datasets to predict what will produce a desired outcome—whether user engagement or strong binding affinity and biological activity. ByteDance’s infrastructure, built to serve billions of users, now supports the compute-intensive demands of training large biomolecular models.
This move aligns with a growing trend of tech companies entering biotech. DeepMind’s Isomorphic Labs, for example, has secured major partnerships with pharmaceutical giants, while other AI-native firms pursue similar generative approaches.
Promise and Remaining Hurdles
Anew Labs is still in the early stages. The IL-17 program remains preclinical, and the company is scheduled to exhibit at upcoming industry events like the BIO International Convention. No clinical data has been released yet.
The broader AI drug discovery field has advanced rapidly, with over 170 programs now in clinical development. Yet the industry’s high failure rate persists—molecules that succeed in the lab do not always translate to safe, effective therapies in humans. Anew’s success will ultimately depend on clinical results, regulatory approval, and manufacturing at scale.
ByteDance’s entry adds significant computational firepower and fresh thinking to the pursuit of previously unreachable medicines. Whether its algorithm for molecules can match the impact of its algorithm for content remains to be seen—but the early signals suggest a serious contender in the race to redefine what is druggable.
About The Author
