Biotechnology · global
After AI Design, How Is It Validated? Tsingke Launches Batch Screening Workflow for Proteins and Antibodies
Much work remains between computer-generated sequences and experimental confirmation of binding. Tsingke links synthesis, expression, and analysis in a two-stage workflow, but daily throughput and delivery speed do not yet reveal how often designs succeed.
Computers can propose hundreds or thousands of protein sequences, but laboratories still need to answer, one by one: Can they be produced? Can they bind to the intended target? Beijing Tsingke Biotechnology (Tsingke) announced a high-throughput validation workflow on September 22, 2026, seeking to bring experimental testing capacity closer to the speed at which AI designs candidate molecules. The service addresses a screening bottleneck in early research; the announcement has not provided evidence of drug candidates’ efficacy.
According to the company’s announcement, the workflow first uses batch gene synthesis and parallel expression to turn designed sequences into testable proteins or antibodies, followed by initial screening using methods such as enzyme-linked immunosorbent assay (ELISA). Researchers can narrow the candidate pool before committing more resources to molecules that perform better in preliminary testing, reducing the burden of conducting in-depth analyses of every design from the outset.
Candidates that pass screening can proceed to scaled-up expression, purification, and gram-scale production, with binding characteristics measured using biolayer interferometry (BLI) or surface plasmon resonance (SPR). These analyses can provide the equilibrium dissociation constant KD, as well as the association and dissociation rate constants kon and koff, helping researchers understand the strength and speed of a molecule’s binding to its target beyond the initial screening signal.
Tsingke claims that its antibody workflow can process up to 1,500 candidates per day and supports single-chain antibody fragments scFv, single-domain antibodies VHH, Fab fragments, and full-length IgG. For qualifying projects, the gene-to-antibody timeline can be as short as seven calendar days, plus five days for shipping. AI Brief’s coverage of the same announcement also retains this eligibility condition; the timeline cannot be taken to mean that all projects can complete gram-scale production and all binding analyses within seven days.
However, throughput does not equal success rate. Industry publication Peptide and Protein News noted that the announcement did not disclose the proportion of designs that succeed or protein expression yields, nor did it name any customers. AI Brief explicitly identified the company’s press release as its information source, so multiple outlets reporting the same figures still cannot be regarded as independent validation of the service’s performance.
These gaps affect how researchers assess the service’s value: if many sequences fail to yield enough testable protein, nominal throughput may not translate into an equivalent volume of usable experimental data. The company also claims it can synthesize DNA fragments up to 200 kb long, but the announcement did not provide delivery timelines or success rates for such long fragments. This claim should not be directly linked to the antibody service’s fastest timeline of seven days.
For AI protein design, the workflow’s potential use is to put more candidates through experimental screening and obtain data for the next round of design. However, confirming that a molecule can bind to a target answers only part of the development question; progressing toward therapeutic use still requires evidence of biological function, safety, and benefit in humans. This announcement outlines the scale and workflow of the validation service, but project-level data are still needed to establish how much it can improve research and development efficiency.