TikTok’s Algorithm Is Now Predicting How Drugs Work.
The company that built TikTok’s recommendation engine, the system that predicts what you want to watch next before you know yourself, now applies the same class of AI to molecules. ByteDance’s drug discovery unit, Anew Labs, stepped into public view in mid-April 2026 with results that the pharmaceutical world did not see coming from a social media company.
From TikTok Feeds to Drug Formulas
Anew Labs presented its first AI-designed therapy at the American Association of Immunologists’ annual meeting in Boston. The molecule targets IL-17, a protein-protein interaction that the pharmaceutical industry has spent decades calling undruggable. The binding surfaces are too large and too flat for conventional small molecules to disrupt. Anew Labs says its AI found a way in.
The unit operates from Shanghai, Singapore, and San Jose, California. Its team of 36 core members works alongside a scientific advisory board that includes Liu Yongjun, former president of Innovent Biologics, Ji Ma, former principal scientist at Amgen, and Hua Zou, scientific director of protein chemistry at Takeda California. The expertise on that board reflects a deliberate strategy: bridge the gap between big-tech computing power and traditional pharmaceutical science.
A High-Stakes Race for AI-Driven Medicine
Beyond the drug presentation, Anew Labs published AnewOmni at the same conference. The framework was trained on 5 million biomolecular complexes. It claims to be the first system to design functional molecules across all biological scales. If clinical validation follows, Anew Labs gets a platform capability, not a single drug program. A repeatable method for finding new drugs is worth far more than any individual compound.
The competition is moving fast. Isomorphic Labs, the DeepMind spinoff backed by Eli Lilly and Novartis, released a drug design tool in February that doubles the accuracy of AlphaFold 3 and holds partnership agreements worth up to $3 billion in milestones. Anthropic acquired Coefficient Bio for $400 million. ByteDance enters a race that well-funded players are already running hard.
What ByteDance brings is a specific organisational muscle. TikTok’s recommendation engine models human behaviour by processing large volumes of data and predicting which content combinations elicit the desired response. Anew Labs’ generative models process enormous volumes of molecular data and predict which atomic combinations produce the desired biological response. The math differs. But the capability, training large models on massive datasets and iterating fast, transfers directly.
Anew Labs has four preclinical programs targeting immunological diseases with large patient populations. The unit exhibits at the BIO International Convention in San Diego in June. The head of computational chemistry presents at the Free Energy Workshop in Barcelona in May. ByteDance built this quietly. It is now ready to be loud.
