Gang Liu at ARDD2026: RushData: High-Throughput Experimental Data for AI-Driven Antibody Discovery

Gang Liu at ARDD2026: RushData: High-Throughput Experimental Data for AI-Driven Antibody Discovery

by Gang Liu · View the channel on PSI

Life & Natural Systems Biology PSALife sciences: general issuesPSAXComputational biology
🎙 Gang Liu 👥 9K 📅 October 23, 2026 ⏱ 20 min 👁 8 📄 expert opinion 🧭 2026-10-23
Available in: English (current) Français

Keywords

RushDataantibodyAIhigh-throughputvalidation

Summary

In this presentation at the 13th Aging Research and Drug Discovery Meeting, Gang Liu from BioIntron introduces RushData, a service designed to generate high-throughput experimental data for AI-driven antibody discovery. He argues that while AI models have made generating thousands of antibody sequences feasible, wet-lab validation has become the new bottleneck. RushData addresses this by testing approximately 3,000 sequences per batch in CHO-K1 cells, delivering supernatant data within 7 days and purified sample affinity and developability data within 14 days. This capability is supported by a factory that produces more than 150,000 antibodies annually. The data output is structured and machine-readable, enabling seamless integration into AI models for iterative learning. Liu emphasizes that the service combines scale, speed, sufficiency (multidimensional data), and structured output, moving beyond traditional fragmentary methods. He also discusses how the facility’s automation and digital infrastructure ensure data quality and reproducibility. In the Q&A, he explains methods for evaluating antibodies when the target is poorly characterized, including thermal stability, self-interaction, and non-specific binding measurements. The talk is informative but serves primarily as a commercial pitch, highlighting a practical solution to a pressing challenge in modern drug discovery.

190 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of the presented information lies in explicitly identifying the bottleneck of wet-lab validation in the AI-driven antibody discovery pipeline and proposing a systematic, scalable solution. The argumentation is logically coherent: AI generates abundant candidates, but without rapid and standardized experimental data, the learning loop cannot close. The speaker supports his claims with concrete operational parameters (batch size, timelines, capacities), which adds practical credibility. However, the argumentation relies heavily on the company’s own infrastructure and does not provide comparative benchmarks against alternative approaches or independent validation. The emphasis on structured, machine-readable data is a valuable point, as it aligns with the needs of AI retraining. Overall, the talk offers a clear rationale but is inherently promotional, lacking external corroboration.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is moderate: the speaker presents company-specific data and processes without citing peer-reviewed literature or independent studies. The methods described (CHO-K1 expression, affinity measurements, developability assays) are standard in the field, but no detailed protocols or quality metrics are provided. The title accurately reflects the content, covering the speaker, topic, and context. No external sources are cited within the talk itself; the only references are the conference and company links in the description. The absence of citations to prior work reduces the overall reliability, though the practical details do offer some transparency. The adequation between title and content is strong, with no discrepancies.

240 words

Title / Content Match

The title accurately represents the content: the speaker, venue, and subject are all correctly indicated. It clearly mentions RushData, high-throughput data, and AI-driven antibody discovery.

Quality & Reliability

6/10

The talk is a technical presentation of a commercial service, providing specific operational details but lacking peer-reviewed validation or independent verification. The methodology is described but not deeply evidenced.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Contribution & Novelties

The talk contributes a concrete example of how to bridge the gap between AI-generated antibody sequences and experimental validation through an industrial-scale service. The novelty lies in the integration of high-throughput production, automated workflows, and structured data output specifically designed for AI retraining. This addresses a critical but often overlooked aspect of the AI-driven drug discovery cycle.

Pour aller plus loin :

  • High-throughput screening — Overview of high-throughput methods in biology, relevant to the scale of RushData.
  • Antibody — Background on antibody structure and function, foundational to the topic.
  • Drug discovery — General context on the drug development pipeline, including the role of validation and AI.

106 words

Radar Profile

The radar profile indicates a relatively high level of technical content and quantity of information, but moderate quality and reliability, reflecting the commercial nature of the presentation. The scores suggest that while the talk is informative and specific, its dependence on company claims and lack of external validation lower its overall trustworthiness.

Reliability 5/10