Insilico Medicine: AI Advances in Pulmonary Fibrosis

Insilico Medicine and Progression to Phase III

Insilico Medicine, a leading company in the use of artificial intelligence for drug discovery, has reached a significant milestone by advancing to Phase III clinical trials. This advancement concerns an innovative drug, identified by AI algorithms, aimed at treating idiopathic pulmonary fibrosis (IPF). This development marks a crucial transition for the sector, moving from initial safety assessments to more advanced validation of therapeutic efficacy.

Idiopathic pulmonary fibrosis is a devastating disease that leads to severe scarring of lung tissue, severely compromising patients' respiratory capacity. Generally, the median survival rate after diagnosis ranges from two to four years. The drug in question, named rentosertib, works by inhibiting the TRAF2- and NCK-interacting kinase, thereby targeting the underlying mechanisms of the disease. This treatment is administered orally, offering a potentially more accessible approach for patients.

A randomized clinical trial was conducted on 71 patients across 22 sites in China. Participants were divided into groups receiving either a placebo or active doses of the drug. The administered doses were 30 mg or 60 mg per day, over a 12-week observation period. Results showed that patients receiving the 60 mg dose per day recorded an average gain in forced vital capacity of +98.4 mL, compared to a loss of 20.3 mL in the placebo group. These results highlight the potential efficacy of the treatment while maintaining manageable safety profiles.

In February 2023, the U.S. Food and Drug Administration (FDA) granted orphan drug designation to rentosertib, recognizing its potential to treat a rare and serious disease.

The Algorithm at the Heart of Discovery

The development of this drug relies on Pharma.AI, Insilico Medicine's proprietary computational pipeline. This system consists of distinct engines that handle specific tasks in biology and chemical engineering. PandaOmics, one of these engines, plays a crucial role in the initial target discovery phase. It analyzes vast biological data sets, including genomics and clinical trial outcomes, to identify patterns in complex biological networks.

PandaOmics successfully isolated TNIK as the primary biological target for intervention in IPF. This approach allowed for bypassing the receptor tyrosine kinase pathways, which are often targeted by existing antifibrotic treatments. The system mapped TNIK as a central node regulating fibrosis and inflammation across various signaling channels, such as Wnt, TGF-β, Hippo/YAP-TAZ, JNK, and NF-κB. By integrating aging markers, PandaOmics assesses the involvement of biological targets in the aging process and chronic inflammation.

Generative Molecular Design

After identifying the target, the Chemistry42 engine takes over for generative molecular design. Unlike traditional high-throughput screening methods, Chemistry42 utilizes tensor reinforcement learning to create molecules that align with the target protein site. This process allows for the design of molecules considering both structural fit and required pharmacological properties.

During this phase, 79 physical molecules were synthesized for testing, and the 55th iteration was selected to advance to preclinical testing. This process significantly reduced the timeline between project initiation and preclinical candidate nomination to just 18 months. The fundamental architecture of this process stems from the publication of the GENTRL methodology by the company in 2019 in Nature Biotechnology, establishing reproducible systems for molecular generation.

Validation through Proteomic Analysis

The clinical evaluation of the drug includes a complex proteomic analysis to validate the biological interactions predicted by the algorithms. Insilico Medicine employs internal proteomic aging clock frameworks to capture exploratory outcomes in geroscience. These clocks, such as ProtAge, OrganAgechrono, ipfP3GPT, and PAOPAC, track changes in biological age resulting from the intervention, using external comparison data sets to contextualize the results.

Research published in Aging and Disease has confirmed that pharmacological inhibition of TNIK produces senomorphic activity, reducing indicators of extracellular matrix remodeling. Signature analyses, such as SenMayo and CellAge, are also conducted to assess the biology of senescence and the senescence-associated secretory phenotype within cellular models.

Documentation and Validation of the Pipeline

The progression of rentosertib through the clinical pipeline is meticulously documented and peer-reviewed, ensuring essential data traceability to verify the capabilities of AI in life sciences. Publications in Nature Biotechnology and the Journal of Medicinal Chemistry detail the various stages of development, from target discovery to clinical validation. Nature Medicine has documented Phase IIa safety data and pulmonary function, providing empirical validation of computational predictions.

Alex Zhavoronkov, founder and CEO of Insilico Medicine, emphasizes that rentosertib represents a pivotal program for the company, illustrating the full arc of their mission. This program demonstrates how AI can not only accelerate the drug discovery process but also open new avenues in therapeutic biology and chemistry.

The ongoing Phase III trial puts generative algorithms to the ultimate test of clinical efficacy, marking a crucial step in the adoption of AI in the biopharmaceutical field.