A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models

Idiopathic pulmonary fibrosis (IPF) is a severe interstitial lung disease with high mortality. Current treatments, nintedanib and pirfenidone, offer limited benefit. The development of new therapies is hampered by challenges in identifying effective drug targets. Early failures in drug development due to poor target selection lead to substantial financial losses and risk aversion by pharmaceutical companies. Data-driven approaches to target discovery, particularly those using AI, are crucial to improve clinical trial success rates. AI has shown promise in identifying targets for other complex diseases, such as those related to aging. Fibrosis, characterized by excessive proliferation of matrix-producing cells, is a common end-stage manifestation of chronic organ failure affecting the lungs, kidneys, and liver. IPF, the most prevalent form of pulmonary fibrosis, is characterized by fibroblast proliferation and extracellular matrix deposition. Myofibroblasts are key players, with transforming growth factor-β (TGF-β) driving their differentiation. The high incidence and cost of IPF treatment represent a major public health challenge. Renal fibrosis is similarly linked to organ failure and represents another unmet clinical need. While TGF-β signaling is central to renal fibrosis, effective antifibrotic therapies remain elusive. The lengthy and costly drug development process, typically exceeding a decade, highlights the need for accelerated approaches. This study uses a generative AI pipeline to address these challenges, focusing on identifying and developing a TNIK inhibitor for fibrosis.

The paper reviews the current limitations in treating IPF and renal fibrosis, highlighting the high attrition rate of drug candidates in clinical trials due to poor target selection. It discusses the use of AI in drug target discovery, citing previous successes in identifying targets for embryonic-fetal transition and muscle aging. The critical role of TGF-β signaling in both IPF and renal fibrosis is emphasized, along with the limited success of existing targeted therapies like nintedanib and pirfenidone. The authors underscore the urgent need for novel anti-fibrotic therapies, highlighting the global burden of these diseases.

The study employed a multi-pronged AI-driven drug discovery pipeline. First, the PandaOmics platform was used to identify potential drug targets for fibrosis. This platform integrates multiple AI engines and analyzes multi-omics datasets from IPF patients, incorporating biological network analysis and text mining of scientific literature. The approach uses a 'time machine' validation strategy, training models on past data and evaluating their ability to predict targets that emerged later in pharmaceutical development. The platform generated a ranked list of targets based on various scores reflecting network connectivity, causal inference, pathway enrichment, and other factors. TNIK was identified as the top-ranked target among protein and receptor kinases, demonstrating connections to fibrosis-driving pathways such as WNT, TGF-β, Hippo, JNK, and NF-κB signaling. Subsequently, the Chemistry42 structure-based drug design platform was used to design small-molecule TNIK inhibitors. Leveraging available crystal structures of the TNIK kinase domain, the AI generated compounds targeting the ATP-binding site and adjacent allosteric pockets. Lead optimization prioritized improved ADME properties, resulting in INS018_055. The compound's binding affinity and selectivity were evaluated using various in vitro assays. The anti-fibrotic activity of INS018_055 was tested in multiple in vitro models, including assays measuring TGF-β-induced EMT and FMT in lung fibroblast and epithelial cells. In vivo efficacy was assessed using murine and rat models of lung and kidney fibrosis, employing different administration routes (oral, inhalation, topical). Finally, the safety and tolerability of INS018_055 were evaluated in two phase I clinical trials, one in New Zealand (NCT05154240) and another in China (CTR20221542), both of which were randomized, double-blinded, placebo-controlled trials involving healthy participants. Detailed information about the protocols for all in vivo and clinical trials is provided in the supplementary information. The study also describes methods used for sample collection and analysis, including Western blotting, immunohistochemistry, ELISA, and high-content analysis, and the specific antibodies, ELISA kits and other reagents that were used in the research.

PandaOmics, an AI-powered target discovery platform, identified TNIK as the top-ranked anti-fibrotic target. The AI-designed TNIK inhibitor, INS018_055, exhibited nanomolar binding affinity and good selectivity in vitro. INS018_055 effectively ameliorated TGF-β-induced EMT and FMT in lung fibroblasts and epithelial cells in vitro. In vivo studies demonstrated that INS018_055 attenuated bleomycin-induced lung fibrosis in mice and rats, with significant improvements in lung function. Oral, inhaled, and topical administrations of INS018_055 showed efficacy. Furthermore, INS018_055 mitigated kidney fibrosis in a murine model. The compound also showed anti-fibrotic activity against dermal fibroblasts in vitro and in a bleomycin-induced skin fibrosis model in rats. Phase I clinical trials in New Zealand and China confirmed the safety and tolerability of INS018_055 in healthy volunteers, showing good oral bioavailability and dose-proportional pharmacokinetics. The complete process from target identification to clinical candidate nomination was completed in approximately 18 months. The study also included transparency analysis of PandaOmics scores and single-cell RNA-seq analysis confirming TNIK's potential role in regulating key cells involved in IPF development.

The study successfully demonstrated the utility of AI-driven drug discovery in identifying and developing a novel anti-fibrotic agent. The rapid development of INS018_055, from target identification to phase I clinical trials in just 18 months, highlights the potential of AI to accelerate the drug discovery process. The findings underscore the role of TNIK in fibrosis and inflammation, two key processes driving IPF progression. The positive results in multiple in vivo models using different administration routes suggest that INS018_055 could be a promising therapeutic candidate for a range of fibrotic diseases. The favorable safety profile observed in phase I trials provides a strong foundation for further clinical development. The broader implications extend beyond IPF, given the lack of effective treatments for other fibrosis-related diseases like renal fibrosis and the potential for preventative use in high-risk populations. The synergy observed with pirfenidone suggests combination therapies might offer additional therapeutic advantages.

This study successfully leveraged generative AI to identify TNIK as a novel anti-fibrotic target and rapidly develop a potent and selective inhibitor, INS018_055. Preclinical studies demonstrated efficacy across multiple fibrosis models, and phase I trials confirmed safety and tolerability. These findings strongly support further clinical investigation of INS018_055 as a potential treatment for IPF and other fibrotic diseases. Future research should focus on phase II and III clinical trials to confirm efficacy in patients and explore potential combination therapies to enhance treatment outcomes.

The study focused primarily on preclinical models and phase I clinical trials in healthy volunteers. Further research is required to confirm efficacy and safety in patients with IPF and other fibrotic diseases. The current study did not address potential long-term effects of INS018_055. More research is needed to fully characterize the long-term effects and potential side-effects associated with prolonged use.