KRAS mutations are notoriously implicated in multiple forms of aggressive cancers, including pancreatic, colorectal, and lung cancers. As a small GTPase, KRAS is pivotal in controlling cellular proliferation and survival pathways, and its hyperactivation due to mutation leads to uncontrolled tumor growth. Historically, KRAS has been deemed “undruggable” because of its extremely high affinity for GDP and GTP nucleotides and the absence of well-defined druggable pockets, hampering the development of effective inhibitors. Insilico Medicine’s breakthrough presents a compelling solution to this long-standing challenge by employing innovative AI-driven drug design methodologies that transcend traditional trial-and-error approaches.

The research commenced by evaluating an existing molecule known for selective inhibition against a particular KRAS variant. Leveraging this molecular baseline, the team sought to identify structurally novel compounds capable of inhibiting all prevalent KRAS mutants—hence the term pan-KRAS inhibitors. Utilizing the generative modules integrated into Chemistry42, researchers synthesized a diverse virtual compound library characterized by various central chemical cores. This diversity was critical for sculpting a new chemical space capable of binding to KRAS in unique and efficacious ways, transcending the limitations of previously known inhibitors.

Scaffold hopping, a strategy to exchange the central core structure of a molecule while retaining or enhancing activity, was rigorously employed through Chemistry42’s virtual screening capabilities. This AI-powered process allowed the team to efficiently explore a vast chemical space, identifying novel core structures that maintain favorable interactions with KRAS binding sites. Concurrently, a suite of molecular modeling and structure-activity relationship (SAR) analyses refined candidate molecules, iteratively optimizing their affinity, selectivity, and pharmacokinetic parameters. Such detailed computational analysis formed the backbone of candidate selection before advancing to physical synthesis.

Once promising molecules were identified, synthesis protocols were established, followed by meticulous biological evaluations. The candidates were tested for their inhibitory potency against multiple KRAS mutants and compared against wild-type KRAS to ascertain selectivity profiles. Encouragingly, the hit series demonstrated a mild selectivity skew towards mutant KRAS variants, exhibiting up to a 4-fold difference in potency, which is a meaningful threshold to minimize off-target effects on normal cellular function. Additionally, the compounds showcased robust inhibition in KRAS mutant cell lines, a crucial preclinical indicator of therapeutic potential.

An often-overlooked hurdle in early drug discovery is the metabolic profile of candidate molecules. Insilico’s research addressed this by assessing cytochrome P450 (CYP) inhibition, a critical determinant of drug-drug interactions and overall drug safety. The pan-KRAS inhibitors displayed acceptable CYP inhibition profiles at this investigative stage, underlining their potential for favorable pharmacodynamics and reduced toxicity risks, which are essential for progressing towards clinical development. This balanced optimization of efficacy and druggability parameters epitomizes the power of AI-enabled drug discovery pipelines.

The fusion of artificial intelligence and human expertise remains at the heart of this scientific triumph. Alex Zhavoronkov, PhD, Founder, and CEO of Insilico Medicine, expressed enthusiasm regarding the transformative potential of the Chemistry42 platform. He emphasized how the integration of advanced molecular modeling and scaffold hopping techniques has facilitated the tackling of KRAS, a target previously considered refractory to drug intervention. This achievement not only validates AI’s role in drug discovery acceleration but also highlights the synergy between computational models and empirical validation.

Insilico Medicine’s journey into AI-driven molecular design dates back to 2016 when the company first introduced the concept of generative AI for novel molecule creation in peer-reviewed literature. This foundational work paved the way for Pharma.AI, a commercial generative AI platform that now spans biology, chemistry, medicinal development, and scientific research. Over the years, Insilico has continuously integrated technological innovations into Pharma.AI, enhancing its capability to innovate rapidly across various fields, including oncology, fibrosis, immunology, pain management, and metabolic disorders.

Beyond oncology, Insilico Medicine applies their AI-driven discovery processes to a broad spectrum of diseases and sectors. Their cutting-edge automated laboratories and in-house drug discovery capabilities enable effective translation of AI-generated candidates into tangible preclinical and clinical assets. Furthermore, the company extends the utility of Pharma.AI beyond healthcare, venturing into advanced materials science, agriculture, nutrition, and veterinary medicine, demonstrating the versatility and scalability of AI in scientific innovation.

The published results of this research, appearing in ACS Medicinal Chemistry Letters, delineate a promising roadmap for future pan-KRAS therapeutics. By embracing a novel chemotype paradigm and coupling it with robust generative and structure-based design methods, Insilico has set the stage for accelerated clinical candidate development against KRAS-driven malignancies. The comprehensive strategy amalgamates AI’s capability to interpret and generate chemical structures with human insight into biological systems and medicinal chemistry, heralding a new era in targeted cancer drug discovery.

In summary, Insilico Medicine’s innovative use of Chemistry42 and generative AI technologies has culminated in the discovery of potent pan-KRAS inhibitors characterized by unique chemical scaffolds, promising selectivity, and favorable drug metabolism profiles. This milestone redefines the potential of tackling “undruggable” targets through AI-enhanced drug design, providing hope for new therapeutic options against cancers with unmet medical needs. As AI continues to evolve and integrate deeper into the drug discovery pipeline, breakthroughs like this exemplify its capacity to transcend traditional pharmaceutical challenges and accelerate the fight against complex diseases.

Subject of Research: Development of novel pan-KRAS inhibitors using AI-driven generative chemistry.

Article Title: Identification of novel pan-KRAS inhibitors via Structure-Based drug design, scaffold hopping, and biological evaluation.

Web References:

• https://pubs.acs.org/doi/10.1021/acsmedchemlett.5c00080
• http://dx.doi.org/10.1021/acsmedchemlett.5c00080

References:
[1] Aladinskiy, V. et al. (2025) "Identification of novel pan-KRAS inhibitors via Structure-Based drug design, scaffold hopping, and biological evaluation," ACS Medicinal Chemistry Letters [Preprint].

Image Credits: Insilico Medicine

Keywords

Generative AI, Oncogenes, Molecular Targets, Medicinal Chemistry

Since the pioneering launch of PandaOmics and Chemistry42 in 2020, Insilico Medicine has consistently showcased the expanding capabilities of its generative AI platforms across various stages of drug development. PandaOmics, a precision target discovery engine, uniquely combines vast omics datasets with sophisticated machine learning algorithms, enabling accelerated identification of disease-relevant molecular targets. Meanwhile, Chemistry42 leverages generative chemistry methods augmented by latest retrosynthesis capabilities to design novel molecules with high specificity and synthetic accessibility, thereby shortening the medicinal chemistry cycle.

The forthcoming Pharma.AI Day will offer an in-depth presentation from Alex Zhavoronkov, PhD, Founder and CEO of Insilico Medicine, detailing the sophisticated advancements underpinning the platform’s newest features. Among these is the integration of single sign-on (SSO) authentication and enhanced genetic data support within PandaOmics, which facilitates streamlined access and broadens the utility of complex genomic inputs for target identification. These enhancements address critical bottlenecks in data interoperability and security, advancing the platform’s role as a comprehensive drug discovery solution.

On the protein engineering front, the Generative Biologics module has received substantive updates to bolster peptide generation and optimization processes. This augmentation refines the platform’s ability to navigate and sculpt the vast biochemical landscape of peptides and proteins, harnessing generative models capable of proposing innovative biologic candidates with potent therapeutic potential. This capability is crucial in the context of biologics, where subtle alterations in amino acid sequences may profoundly influence efficacy and immunogenicity.

Complementing these software advancements is Life Star1, Insilico Medicine’s sixth-generation automated laboratory system. This AI-driven intelligent robotics lab exemplifies the seamless amalgamation of computational predictions and empirical validation. By automating iterative synthesis, screening, and data collection, Life Star1 dramatically accelerates the experimental feedback loop essential for optimizing drug candidates. The platform’s continual evolution signals a future where AI-generated hypotheses are rapidly corroborated or refined in fully integrated wet lab environments.

Central to Insilico Medicine’s innovation pipeline are the Large Language of Life Models (LLLMs), known under the PreciousGPT series. Since the debut of Precious3GPT in mid-2024, these models have undergone fine-tuning and expansion, augmenting their proficiency in natural geroprotector discovery and automated compound screening. By leveraging transformer-based architectures customized for biological context, PreciousGPT exemplifies how domain-specific language models can unravel complex biochemical patterns and propose viable therapeutic interventions targeting aging and age-related pathologies.

The generative chemistry suite benefits from the enhanced capabilities of Retrosynthesis, a synthetic route prediction engine now embedded within Chemistry42. Retrosynthesis facilitates forward and backward design of chemical molecules by evaluating feasible synthetic pathways, thus furnishing medicinal chemists and AI agents with pragmatic blueprints for molecule production. Further enriching this capability is Nach01, a foundational multimodal model trained on both natural and chemical languages, representing a novel frontier in integrating disparate data modalities for holistic drug design.

Science42: Dora, a versatile AI agent designed for scientific writing assistance, will also be spotlighted during Pharma.AI Day. This tool incorporates expanded document template libraries and advanced AI integrations, elevating the efficiency of scientific communication. By automating literature synthesis, experiment planning, and manuscript drafting, Dora empowers researchers to focus on innovation, reducing administrative burdens and accelerating knowledge dissemination.

Insilico Medicine’s journey began with the seminal publication in 2016 that introduced the concept of generative AI for novel molecule design. This groundbreaking work underpinned the commercial rollout of the Pharma.AI platform, which has since facilitated the nomination of over 22 developmental and preclinical candidates across diverse therapeutic areas, from fibrosis to oncology. Impressively, internal programs have achieved average timelines of 12 to 18 months to developmental candidate stage, synthesizing and testing between 60 to 200 molecules per program, showcasing the platform’s throughput and precision.

The company’s AI-driven pipeline portfolio boasts ten molecules with Investigational New Drug (IND) clearances. Among them, Rentosertib (formerly ISM001-055) stands out as a potential first-in-class treatment for idiopathic pulmonary fibrosis, having successfully completed Phase 2a clinical trials with encouraging safety and efficacy data. This achievement marks a significant milestone in translating generative AI discoveries into tangible clinical advancements.

Moreover, Insilico Medicine is actively extending its AI and automation expertise beyond conventional drug discovery. Current exploratory efforts include breakthroughs in aging research, deploying AI to identify novel geroprotectors; sustainable chemistry initiatives aimed at reducing environmental impact through AI-guided molecular design; and agricultural innovation targeting enhanced crop resilience and productivity. These endeavors underscore the versatility and societal impact potential of the company’s technology portfolio.

Pharma.AI Day 2025 represents not only a showcase of Insilico Medicine’s technological prowess but also a platform for fostering collaboration and open innovation within the pharma and biotech communities. By sharing quarterly updates and live demonstrations, the event facilitates direct dialogue between AI developers, computational biologists, medicinal chemists, and clinical researchers, accelerating the collective effort to overcome longstanding biomedical challenges.

For researchers, practitioners, and enthusiasts keen on the frontier of AI-driven life sciences, registering for the webinar offers an opportunity to witness firsthand the cutting-edge synthesis of computation, automation, and translational science. Insilico Medicine’s continued evolution of Pharma.AI heralds a new era where artificial intelligence is not just a tool but a central architect in the discovery and development of next-generation therapeutics.

—

Subject of Research:
Generative Artificial Intelligence Applications in Drug Discovery and Life Sciences Automation

Article Title:
Insilico Medicine Gears Up for Pharma.AI Day 2025, Unveiling Cutting-Edge Advances in AI-Driven Drug Discovery

News Publication Date:
April 18, 2025

Web References:
https://insilico.zoom.us/webinar/register/WN_KQxBpQSaQzeWh3O6WfbITA#/registration
https://pharma.ai/pandaomics
https://pharma.ai/generativebiologics
https://pharma.ai/chemistry42
https://pharma.ai/science42/dora
http://insilico.com

Image Credits:
Insilico Medicine

Keywords:
Generative AI, Drug Discovery, Biological Models, Molecular Targets, Medicinal Chemistry, Clinical Research, Genetic Screening