AI+ Pharma has greatly improved the efficiency and success rate of new drug R&D

Chinanews.com, Shanghai News, October 9 (Reporter Chen Jing) The R&D investment of new drugs is large, the cycle is long, and the risk is high, according to Nature data, the average R&D cost of a new drug is about 2.6 billion US dollars, it takes about 10 years, and the success rate is less than 10%. In recent years, the development of artificial intelligence (AI) has brought new technical means to new drug research and development, which is expected to be applied to multiple scenarios and stages in drug research and development, helping to improve the efficiency and success rate of new drug research and development.

The reporter learned on the 9th that Insilico Medicine not only proposes a new paradigm to transform early drug discovery with generative AI as the core, but also breaks the traditional AI method of predicting results based on existing data, and proposes a solution for artificial intelligence to penetrate the R&D process in an all-round way to reduce costs and increase efficiency.

Driven by generative AI, the world's first "AI drug" has been validated for the first time in the patient population, which is a major progress in the global AI pharmaceutical industry. For the 5 million patients around the world who are at risk of death, the rapid entry into the critical phase of Phase 2 clinical trials is expected to be the dawn of overcoming this intractable disease.

It is understood that in 2016, Insilico Medicine published a paper that applied breakthroughs in cutting-edge technologies such as generative adversarial networks (GANs) to drug discovery for the first time, and in the following three years, Insilico Medicine has experienced the transition from the era Pharma.AI of algorithms to the era of software, and has integrated many optimization models such as the generative tensor reinforcement learning model (GENTRL) published in Nature Biotechnology in its self-developed drug development platform.

Insilico Medicine takes the world's first AI-powered drug INS018_055 that uses AI to discover novel targets, design innovative molecular structures, and successfully enter Phase 2 clinical trials as an example, combined with a recent scientific paper published in Nature Biotechnology, to explain in detail the AI-accelerated multi-module generative AI workflow for target discovery, drug development, and clinical trials.

INS018_055 targets the deadly rare disease Idiopathic Pulmonary Fibrosis (IPF), which is characterized by impaired lung function due to fibroblast proliferation and massive extracellular matrix deposition, with a median survival of only 2-3 years after diagnosis. Currently, less than 30% of patients benefit from approved targeted therapies. Professor Xu Zuojun, chief physician of Peking Union Medical College Hospital, Chinese Academy of Medical Sciences, said that idiopathic pulmonary fibrosis, or IPF, is the hardest bone in interstitial lung disease. It is largely ineffective against our conventional glucocorticoid and immunologic treatments. Insilico Medicine's clinical trial is an exploratory study.

To build the initial disease target hypothesis, the R&D team used PandaOmics, the target discovery engine of the Insilico Medicine Pharma.AI platform, to train on age- and sex-annotated omics data and clinical datasets, and then use the iPANDA algorithm published in Nature Communications in 2016 to nominate potential targets through deep feature synthesis, causal inference, and novel pathway reconstruction. After that, the team fused natural language processing (NLP) engines to evaluate novelty and score disease-target correlation based on millions of files covering text data such as patents, publications, R&D funds, and clinical trials. Among the 20 potential targets revealed by the PandaOmics platform, Traf2 and Nck interacting kinase (TNIK) stood out and was finally identified as a key research object.

After the target was identified, the team used Chemistry42, a Pharma.AI-based generative chemistry platform, to generate a safe, specific, and efficient TNIK inhibitor using a structure-based drug design (SBDD) strategy. At the same time, the platform uses 30 generative AI models to design compounds, form a virtual structure library, and receive feedback from professional R&D teams to further optimize the virtual screening process. After multiple screenings, the TNIK ATP binding site was selected as the target binding pocket, and one of the potential lead compounds exhibited excellent activity with IC50 values in the nanomolar range.

Based on the above de novo generation steps, the R&D team carried out further optimization to improve solubility, optimize ADME properties, and reduce toxicity while retaining the strong affinity of the candidate molecule for the TNIK target. INS018_055 has been validated in a variety of animal models of fibrosis, marking the transition of AI-driven drug discovery from theory to reality. In mice and rats that induce pulmonary fibrosis, INS018_055 improve lung function by decreasing fibroblast activation, reducing fibrotic protein deposition, and reducing lung inflammation. INS018_055 also attenuated skin and renal fibrosis in two in vivo models, exhibiting pan-fibrosis inhibition efficacy, indicating potential indication expansion opportunities.

Insilico Medicine subsequently conducted a randomized, double-blind, placebo-controlled Phase 1 clinical trial (NCT05154240) in New Zealand to evaluate the safety, tolerability, and pharmacokinetic properties of INS018_055 in 78 healthy volunteers. , INS018_055 overall safety and good tolerability, with no serious adverse reactions or deaths reported in clinical trials. All treatment-related adverse effects were mild and resolved at the end of the study.

It is reported that AI-guided target discovery and selection is expected to improve the success rate of drug R&D, reduce R&D costs by avoiding wrong target selection and reducing duplication of work. In addition, AI tools such as Chemistry42 will leverage the development of computing power to further simplify the process of generating innovative small molecules and drive revolution in drug discovery.

During this R&D process, the Chinese team undertook the vast majority of drug development tasks. Feng Ren, Co-CEO and Chief Scientific Officer of Insilico Medicine, said that at the beginning of the project, it was Insilico Medicine's global AI team that first proposed the target hypothesis through PandaOmics and screened the molecule through Chemistry42, and the Chinese drug development team intervened at the stage of identifying hit compounds. In February 2021, we announced the nomination of a preclinical candidate for this project, which was achieved by the Chinese team and the global team.

Professor Xu Zuojun said that in the future, we will further communicate with the Insilico Medicine team and the regulator to improve and adjust the clinical protocol, hoping to replicate the positive results of the Phase IIa clinical trial in a larger IPF patient population. At the same time, we look forward to seeing the opportunity for this project to break through the approval review. "Artificial intelligence is increasingly being used in all aspects of clinical research. It includes disease diagnosis, disease treatment, prognosis judgment, and drug development. I firmly believe that AI will certainly play an increasingly important role in this regard. He said. (ENDS)

Editor: Chen Jing