Generative AI aids discovery of ISM3830's novel scaffold

Casitas B-lineage lymphoma-b (CBLB) is an E3 ubiquitin–protein ligase that functions as an intracellular checkpoint and master negative regulator of T‑cell and natural killer (NK) cell activation, playing a central role in modulating T-cell activation and immune tolerance, and its inhibition enhances T cells and NK cells function, also provides a strategy for the function restoration of exhausted T cells.

According to previous studies, CBLB is highly expressed in many immune subsets across multiple cancer types, indicating its great potential as a therapeutic target for cancer immunotherapy, especially in advanced colorectal cancer, prostate cancer, renal carcinoma, melanoma, etc. 

Insilico Medicine ("Insilico"), a clinical-stage, generative AI-driven drug discovery company, today announced the nomination of ISM3830, a potential best-in-class, oral available, highly selective CBLB inhibitor with AI-empowered novel scaffold, as a preclinical candidate against multiple advanced tumors.

"With repeated real-world proof including preclinical candidate nomination and positive clinical results, we are growing even more confident in generative AI's potential to accelerate drug development and enable true therapeutic innovation", said Feng Ren, Ph.D., Co-CEO and CSO of Insilico Medicine.

Equally important, CBLB inhibition, from its fundamental mechanism of action, supports indications with low response or resistance to current immune checkpoint inhibitors-an area of significant unmet need that we are determined to address with the combined force of AI speed and human intelligence."

Feng Ren, Ph.D., Co-CEO and CSO, Insilico Medicine 

It is worth noting that the novel scaffold of ISM3830, with a similarity index to current available molecule of 0.42, was discovered with the aid of Chemistry42 and its ADMET predictor module. Leveraging a structure-based strategy based on currently available cocrystal structures, novel candidate compounds were generated through 40+ generative AI models integrated in Chemistry42, evaluated using ADMET predictor module, and further optimized in repeated generation attempts guided with built-in reward pipelines. Afterwards, synthesis and comprehensive bioassay testing finally yielded ISM3830 as a highly promising drug candidate, with distinctive advantages listed as follows: 

• Empowered by Insilico's proprietary generative AI platform, the drug candidate overcomes the current bottlenecks of CBLB inhibition therapies in metabolism and absorption. 

• Preclinical safety screening demonstrated low risk of hypotension, gastrointestinal toxicity, and off-target toxicity, as well as excellent selectivity and improved safety margin in DRF study. 

• Promising druggability and excellent ADME/PK profile both in vitro and in vivo, measured by lower in vivo clearance, and higher oral bioavailability across preclinical species. 

Preclinical studies suggest robust in vivo efficacy in mice models and induction of long-term tumor immunity, verified by CT26 rechallenge experiments. Additionally, combination potential was with a broad spectrum of treatment options, including immune checkpoint inhibitors, chemotherapies, and other targeted agents. 

Besides the milestone achievement of PCC nomination, Insilico R&D team has recently published the development process of a CBLB inhibitor of the same series, in a peer-reviewed paper on Journal of Medicinal Chemistry, titled "Discovery and Biological Evaluation of Novel, Potent, and Orally Available CBLB Inhibitors", demonstrating how cocrystal structure analysis, iterative structure–activity relationship (SAR) exploration, systematic optimization on potency, selectivity, and pharmacokinetic profiles, combined with structure-based AI generation results could provide a roadmap for further optimization of CBLB inhibitors. 

In March 2024, Insilico published in Nature Biotechnology the journey of the Rentosertib program from inception to Phase 1 clinical trials, along with part of experimental data. In December 2024, Insilico published in Nature Biotechnology the AI-enabled preclinical research journey and partial experimental data for the gut-restricted PHD1/2 inhibitor ISM5411. In January 2025, Insilico, together with partners including the University of Toronto, published research in Nature Biotechnology on exploring generative AI with a quantum-classical hybrid model to design novel KRAS inhibitors. In May 2025, Insilico published collaborative research in Nature Communications on AI-enabled development of pan-coronavirus inhibitors. In the same month, Insilico published another Nature Communications paper on AI-enabled development of next-generation ENPP1 inhibitors for innate immune modulation.

Notably, Insilico's June publication in Nature Medicine marked the world's first clinical proof-of-concept for AI drug discovery through the success of Rentosertib (ISM001-055), the potential first-in-class drug candidate with both novel target and innovative structure discovered with the help of AI. 

By integrating AI and automation technologies and deep in-house drug discovery capabilities, Insilico is delivering innovative drug solutions for unmet needs including fibrosis, oncology, immunology, pain, and obesity and metabolic disorders, nominating 23 preclinical candidates since 2021, among which 10 molecules have received IND approval.