ByteDance unveils first AI-designed therapy, enters drug discovery race

ByteDance, the company behind TikTok’s eerily accurate recommendation algorithm, is now applying a similar breed of artificial intelligence to a vastly different challenge: designing drugs. Its drug discovery arm, Anew Labs, recently unveiled its first AI-designed therapy at the American Association of Immunologists’ annual meeting in Boston. The presentation detailed a generative AI-designed small molecule that targets IL-17, a cytokine implicated in autoimmune conditions such as psoriasis, rheumatoid arthritis, and ankylosing spondylitis.

This molecule aims to disrupt a protein-protein interaction, a class of drug target long dismissed by the pharmaceutical industry as “undruggable.” The reason? The binding surfaces are typically too large and flat for conventional small molecules to latch onto effectively. Anew Labs claims its AI has cracked this barrier, finding a novel path where traditional chemistry has failed.

The Boston conference marked the first public glimpse of ByteDance’s drug unit capabilities, but it won’t be the last. Anew Labs is already scheduled to exhibit at the BIO International Convention in San Diego this June, and its head of computational chemistry will speak at the Free Energy Workshop in Barcelona next week.

Operating from hubs in Shanghai, Singapore, and San Jose, California, Anew Labs lists 36 core team members on its website. Its scientific advisory board reads like a who’s who of biologics and immunology, including Liu Yongjun, former president of Innovent Biologics; Ji Ma, a former principal scientist at Amgen; and Hua Zou, scientific director of protein chemistry at Takeda California. These experts come from companies where the targets Anew Labs pursues have historically required costly injectable antibody therapies, often costing tens of thousands of dollars annually. The unit’s goal is to replace these injections with oral pills, leveraging generative AI to create small molecules that mimic antibody functions but in a patient-friendly, swallowable form.

Chris Li, Anew Labs’ head of biology, presented one of the unit’s four pipeline candidates in Boston. This molecule is a pan-spectrum IL-17 inhibitor, designed to block multiple variants of the IL-17 cytokine simultaneously. Existing treatments, such as Novartis’s secukinumab and Eli Lilly’s ixekizumab, are injectable antibodies that generate billions in revenue annually by treating psoriasis and other inflammatory conditions. An oral small molecule achieving comparable efficacy would be commercially transformative, offering cheaper manufacturing and easier administration. However, the challenge remains formidable: IL-17’s binding surface is a broad, shallow interface that offers little for a small molecule to grip. The gap between AI’s lab performance and real-world patient outcomes is the defining tension in health tech, and IL-17 sits squarely at the center of that divide.

In March, Anew Labs published a preprint on bioRXiv describing AnewOmni, a generative AI framework trained on over five million biomolecular complexes. This model is designed to operate across molecular scales, from small chemical compounds to peptides to nanobodies, assembling chemically meaningful building blocks at atomic resolution. The preprint demonstrated that AnewOmni could design functional molecules targeting KRAS G12D, a heavily studied oncology target, and PCSK9, a cholesterol-related protein, achieving success rates between 23 and 75 percent with only low-throughput laboratory validation. The model uses programmable graph prompts, allowing researchers to guide the generation process by specifying chemical, geometric, and topological constraints.

The technical breakthrough here is significant because most generative models excel at one molecular scale but fail when asked to design across scales. A model that works for small molecules typically cannot also design peptides or protein-based therapeutics. AnewOmni claims to be the first framework to succeed at functional molecular design across all scales. If validated in clinical settings, this would give Anew Labs a platform capability rather than a single-programme advantage. Competitors like Isomorphic Labs, the DeepMind spinoff backed by Eli Lilly and Novartis, released its own drug design tool in February that doubles AlphaFold 3’s accuracy, with partnership agreements worth up to $3 billion in milestone value. The race to build the definitive AI drug design platform is global, and ByteDance has entered it with a model that, on paper, addresses a limitation its rivals haven’t yet publicly solved.

ByteDance is not the first tech giant to pivot into drug discovery. Anthropic acquired Coefficient Bio for $400 million in an acqui-hire that brought fewer than ten people into its biological research efforts. Google’s DeepMind has focused on protein structure prediction since AlphaFold’s breakthrough, which won the 2024 Nobel Prize in Chemistry. Microsoft has invested in biology-focused AI through its partnership with Paige, a computational pathology company. Nvidia built BioNeMo, a platform for training and deploying biomolecular AI models. The pattern is consistent: companies with the most advanced AI infrastructure are redirecting that capability toward biology because drug discovery is a problem shaped like the ones AI excels at,searching vast combinatorial spaces for rare solutions that satisfy multiple constraints simultaneously.

What distinguishes ByteDance’s entry is the source of its AI expertise. TikTok’s recommendation engine is fundamentally a system that models human behavior by processing enormous data volumes and predicting which content combinations will produce the desired response. Anew Labs’ generative models do something structurally similar: they process massive amounts of molecular data and predict which atomic combinations will produce the desired biological response. The mathematical architectures aren’t identical, but the organizational capability,training large models on massive datasets, iterating rapidly, and deploying at scale,is transferable. ByteDance’s AI infrastructure, built to serve 1.5 billion TikTok users, is now applied to a problem where the “users” are molecules and the “engagement metric” is binding affinity.

More than 173 AI-discovered drug programmes are now in clinical development globally, with 15 to 20 entering large-scale trials this year. Whether AI will revolutionize drug development depends on how it’s used, and the industry’s 90 percent clinical failure rate hasn’t yet demonstrably improved. Insilico Medicine’s rentosertib, a first-in-class drug for idiopathic pulmonary fibrosis where both the target and molecule were AI-discovered, showed positive Phase IIa results published in Nature Medicine. The Recursion-Exscientia merger created the most comprehensive AI drug discovery platform in the industry, but then discontinued its lead AI-discovered candidate after long-term data didn’t confirm earlier efficacy trends. The pattern across the field is promising early data followed by the same biological reality that has always made drug development difficult: molecules that work in a dish don’t always work in a body.

Anew Labs has four pipeline candidates and a generative platform that, if its preprint results hold, can design functional molecules across scales. It has the backing of a parent company valued at roughly $300 billion with AI infrastructure that dwarfs most pharmaceutical companies’ computational resources. It has advisors from Innovent, Amgen, and Takeda. What it does not yet have is clinical data. The IL-17 molecule presented in Boston was preclinical. The distance from a poster at an immunology conference to an approved oral therapy that replaces injectable antibodies is measured in years and billions of dollars, and most molecules that start that journey don’t finish it. The most ambitious AI-biology startups understand that the algorithm is the beginning, not the end. ByteDance built an algorithm that changed how a billion people consume content. Whether the same company can build an algorithm that changes how a disease is treated is a question no conference presentation can answer. Only a clinical trial can.