A new study published in Engineering has combined machine learning (ML) and experimental validation to identify dihydromyricetin (DHM), a natural flavonoid, as a potent inhibitor of the TGF-β/ALK5 signaling cascade with therapeutic potential for idiopathic pulmonary fibrosis (IPF). IPF is a chronic, progressive, and life-threatening lung disorder marked by diffuse alveolitis, structural disruption, and excessive extracellular matrix deposition, with a median survival of only 3 to 5 years. Current pharmacotherapies, including pirfenidone and nintedanib, only modestly slow disease progression and cannot halt or reverse fibrosis, creating an urgent need for novel therapeutic approaches.

In this work, researchers built an ML classification model fine-tuned from the Uni-Mol framework to predict inhibitors of the canonical TGF-β/SMAD pathway from natural products. The model was trained and validated using affinity data from BindingDB and externally tested with ChEMBL compounds, achieving strong performance with an area under the precision–recall curve (AUPRC) of 0.936 and an area under the receiver operating characteristic curve (AUROC) of 0.902. The team screened a library of more than 16 700 herbal natural products and predicted 408 potential inhibitors, from which 20 were selected for experimental confirmation. Using a dual-luciferase reporter assay, DHM emerged as the most potent candidate that suppressed the TGF-β/SMAD signaling cascade.

In vitro experiments showed that DHM suppressed TGF-β1-triggered epithelial–mesenchymal transition (EMT) in A549 cells and fibroblast transdifferentiation in MRC-5 cells, while also reducing cell migration induced by TGF-β1. In a bleomycin-triggered mouse model of pulmonary fibrosis, DHM attenuated fibrotic lesions and inflammatory responses, improved respiratory function indicators, and lowered collagen accumulation and hydroxyproline levels in lung tissues. Mechanistic investigations revealed that DHM targets the type I TGF-β receptor, also known as ALK5, directly binds to the receptor’s kinase domain, reduces its membrane expression, and represses its kinase activity, leading to downregulation of both SMAD-dependent and non-canonical TGF-β signaling pathways.

This research is the first to report DHM as a TGF-β/SMAD pathway inhibitor identified via ML with demonstrated efficacy against IPF. The anti-fibrotic activity of DHM is mediated through ALK5 blockade, which suppresses downstream signaling, EMT, and fibroblast activation. Compared with currently available drugs, DHM displays a favorable safety profile and good water solubility, supporting its promise as a candidate for anti-IPF drug development. These findings also underscore the value of integrating computational and experimental methods in discovering natural product-based therapies for complex lung diseases.

The paper “Machine Learning-Enabled Insights: Dihydromyricetin’s Novel Role in Inhibiting the TGF-β/ALK5 Signaling Cascade for the Treatment of Pulmonary Fibrosis,” is authored by Luyao Dong, Wenting Dong, Yixin Ren, Chunjie Xu, Xiukun Wang, Peiyi Sun, Yao Meng, Congran Li, Guoqing Li, Jiandong Jiang, Hao Wang, Xuefu You, Xinyi Yang. Full text of the open access paper: https://doi.org/10.1016/j.eng.2025.10.017. For more information about Engineering, visit the website at https://www.sciencedirect.com/journal/engineering.