AIPher (艾斐智药)

Tsinghua AIR spin-out commercializing DrugCLIP, a genome-scale virtual-screening model, through co-development and an ADHD lead.

1. Core Product / Service

AIPher (艾斐智药) was incorporated in July 2026 and incubated by Tsinghua University's Institute for AI Industry Research (AIR) [1][2]. Its core asset is DrugCLIP, a virtual-screening model:

• Design. A two-tower contrastive-learning model that maps protein pockets and small molecules into one shared embedding space, so screening becomes fast vector search instead of docking [2][6].
• Scale. The company says one server can evaluate about 10 trillion pocket–molecule pairs per day [2].
• Origin and publication. DrugCLIP was developed by the joint health-computing center of Tsinghua AIR and the Beijing Academy of Artificial Intelligence. Its human-genome-scale screening results were published in Science in January 2026, and a free web version was released [6].

On top of DrugCLIP the company is building PharmAgents, an agent layer for the wider discovery workflow [1][7].

Pipeline.

• Lead program: ADHD, based on a newly identified allosteric pocket on the norepinephrine transporter (NET). It is at pre-PCC / PCC confirmation [2][3].
• Co-development: active projects with five or more pharma and biotech partners in metabolic disease, CNS, autoimmunity and drug delivery [2].
• CACHE challenge: in this blind hit-finding challenge, the molecules the team picked were the only ones, out of more than 1,300 submissions, active in all three experimental assays. The organizers synthesized and tested them, so this is independent validation [3].

Wet-lab model: no in-house lab disclosed. Validation relies on academic partners. Reported partners include Tsinghua's School of Life Sciences and School of Pharmacy, the Chinese Academy of Agricultural Sciences and a Shanghai Jiao Tong translational-medicine school [2][4]. Whether CROs are used is not disclosed. This puts AIPher at the "light wet-lab" end of Chinese AI-native startups.

2. Target Users & Pain Points

• Pharma and biotech partners who want hits against hard or new pockets fast. Screening billions of compounds against genome-scale targets is too expensive with docking, and DrugCLIP's speed is the selling point.
• Its own CNS program. Finding allosteric sites on transporters is the kind of problem where pocket-level representation learning may help.
• Business model: platform services and co-development in the short term, with its own assets later [2].

3. Competitive Landscape

4. Unique Observations

• The team is older than the company. The team has worked on AI drug discovery since 2021, and the DrugCLIP Science paper predates incorporation. "Founded 2026" says little about maturity.
• Its strongest external evidence is unusually clean. Most startup claims are self-reported. The CACHE result was produced by third-party synthesis and assays, which is stronger evidence than typical press-release hit rates, though it shows hit-finding skill, not drug-development ability.
• Relying on academic labs keeps costs low but caps speed. Without its own lab or a dedicated CRO setup, design–test cycles depend on partner schedules. That works for hit discovery but becomes a bottleneck at lead optimization and IND-enabling stages.
• There is a sibling company. ligit-pharma was founded months later by another DrugCLIP co-first author. Whether the two have an equity, licensing or competitive relationship is not public.

5. Financials / Funding

• Angel round (Sept 2026): about ¥100M (RMB) [1][2][7]. Closed roughly one month after incorporation.
  • Lead: Xianghe Capital (襄禾资本).
  • Followers: Xinglian Capital (星连资本, affiliated with Zhipu) and an unnamed industry fund.
• Use of funds: DrugCLIP and PharmAgents development, and pipeline progress [7].
• Revenue: not disclosed.

6. People & Relationships

• CEO: 高博文 (Gao Bowen). Former recommendation-algorithm engineer at ByteDance; returned to Tsinghua AIR for a PhD in 2022; co-first author of the DrugCLIP Science paper [2][3].
• Chief Scientist: 兰艳艳 (Lan Yanyan), Vice Dean of Tsinghua AIR [2].
• Chief Scientific Advisor: 张亚勤 (Zhang Ya-Qin), Dean of Tsinghua AIR [2].
• Other core members: an R&D head with medicinal-chemistry and structural-biology background, and a business lead with nearly 20 years of pharma experience, both unnamed [2].
• Institutional origin: Tsinghua AIR. An AstraZeneca–Tsinghua joint research center was announced in March 2026; whether it involves AIPher directly is unclear [5].
• Investors: Xianghe Capital, Xinglian Capital (zhipu-affiliated).
• Peers: ligit-pharma, anew-labs, xtalpi, dp-technology.

Sources

1. Tencent News — https://news.qq.com/rain/a/20260903A033WN00
2. 36Kr — https://eu.36kr.com/zh/p/3971434119714441
3. Sina Tech — https://finance.sina.com.cn/tech/shenji/2026-09-03/doc-iniqpeiy4345897.shtml
4. Beijing News — https://www.bjnews.com.cn/detail/1788401845129243.html
5. STAR Market Daily — https://www.chinastarmarket.cn/detail/2473583
6. Xinhua — https://www.news.cn/tech/20260109/2e0f65d6733a4e2588a97dfe96593a09/c.html
7. ChinaVenture — https://m.chinaventure.com.cn/news/111-20260903-393077.html