Challenges in AI Drug Discovery Clinical Success Rates: Insilico’s Lenticelib First to Phase II, AlphaFold Accelerates Structural Biology

July 5, 2026

MDPI Switzerland

Overview

While AI drug discovery accelerates early-stage discovery, consistent improvement in late-stage clinical trial success rates remains a challenge. However, Insilico Medicine’s lenticelib (NCT05938920) has become one of the first AI-assisted drug candidates to enter Phase II clinical evaluation. Furthermore, AI-powered structural prediction tools like DeepMind’s AlphaFold have significantly accelerated structural biology research, adding over 850 new structures to the Protein Data Bank (PDB).

In Depth

AI Drug Discovery: Bridging the Gap Between Early-Stage Acceleration and Late-Stage Clinical Success

Artificial Intelligence (AI) has demonstrably revolutionized the early phases of drug discovery, dramatically shortening timelines for hit identification, lead optimization, and preclinical design. While AI generates numerous promising drug candidates with enhanced efficiency and speed, a consistent improvement in late-stage clinical trial success rates for AI-assisted drugs has yet to materialize, posing a significant ongoing challenge for the field.

Insilico Medicine’s Lenticelib Marks First AI-Discovered Drug in Phase II

Against this backdrop, Insilico Medicine’s lenticelib (ISM001-055, clinical trial registration number NCT05938920) stands as a pivotal milestone: it is among the first AI-discovered and designed drug candidates to advance into Phase II clinical evaluation. This progression provides strong evidence that AI is not merely a research tool but is capable of generating genuine therapeutic candidates for human use, bolstering confidence in the feasibility of AI drug discovery.

AlphaFold’s Transformative Contribution to Structural Biology

Another profound success of AI lies in structural prediction tools, exemplified by DeepMind’s AlphaFold. AlphaFold has solved the long-standing problem of accurately predicting protein 3D structures, leading to the deposition of over 850 new protein structures into the Protein Data Bank (PDB). This has significantly accelerated structural biology research, expanding possibilities for elucidating the function of many previously uncharacterized proteins and for designing drugs that target them effectively.

Underlying Challenges and Context

The initial success of AI in drug discovery stems primarily from advances in computational power and data analytics. However, the high failure rates in clinical trials, especially in later phases, are attributable to biological complexity, the inherent difficulty of prediction, and stringent criteria for clinical safety and efficacy. While AI can optimize molecular design, fully predicting drug behavior within complex in vivo systems and individualized patient responses remains challenging. Addressing this gap requires further refinement of AI models and the integration of more comprehensive biological and clinical datasets.

Future Outlook

The success of lenticelib’s Phase II trial would further pave the way for AI in clinical applications. Tools like AlphaFold will continue to provide indispensable information for basic research and early drug discovery, enhancing the quality of the pipeline. Moving forward, AI is expected to expand its application to later development stages, including optimizing clinical trial design, patient stratification, and real-world data analysis. The synergistic integration of AI and human expertise will be key to improving success rates across the entire drug discovery process, delivering innovative therapies to patients more rapidly.

Source: https://www.mdpi.com/1424-8247/19/6/916

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.