Identification of characteristic genes and herbal compounds for the treatment of psoriasis based on machine learning and molecular dynamics simulation

Abstract

Psoriasis brings economic and mental burdens to patients, the exact etiology and pathogenesis of psoriasis are still unclear. Compounds of herbal medicine have the potential for psoriasis treatment. This study aims to explore the characteristic genes for psoriasis, which herbal compounds may target. Four differential gene expression datasets, with 181 healthy skin and 181 psoriasis skin lesion samples, were used for analysis. This study employed random forest, neural network, and support vector machine algorithms to identify the characteristic genes associated with psoriasis. The identified genes were validated using external datasets. Then, the main compounds were identified. The targets of compounds were collected through SwissTargetPrediction, Super-PRED, HERB databases, and so on. Finally, a batch virtual screening of compounds with the identified characteristic genes was conducted. Open Babel and AutoDock Tools 1.5.6 were used for molecular docking, and Desmond was used to evaluate molecular dynamics simulations. Twelve characteristic genes, successfully validated in external datasets genes, were identified from 1270 differential genes. The 59 compounds identified contained 1795 targets. There are 143 intersections between differential genes and compound targets. Two-hundred and ninety-four compound-target combinations were selected for molecular docking screening. It was finally found that 8 protein-ligand combinations are highly critical for treating psoriasis, namely AKR1B10-Astilbin, AKR1B10-Ferulic acid, AKR1B10-Cianidanol, IL36G-Astilbin, MMP9-Ferulic acid, OASL-Astilbin, PPARG-Astilbin, SERPINB3-Astilbin, molecular dynamics simulations also indicate that these eight pairs of combinations are stable. This research brings a new perspective to the treatment of psoriasis, these characteristic genes and compounds deserve the attention of clinical researchers.

Communicated by Ramaswamy H. Sarma

Keywords:

• Psoriasis
• machine learning
• bioinformatics
• network pharmacology
• molecular dynamics simulation

Authors contributions

DWC designed the study. DWC, WZX, and LXF conducted data collection. DWC analyzed and interpreted the data. DWC drafted the manuscript. YF, GWY, and LSY provided critical comments, and ZGZ revised the manuscript. All authors read and approved the final version of the manuscript.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The psoriasis datasets analyzed for this study can be found in the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Patent data is available through the China National Intellectual Property Administration (http://epub.cnipa.gov.cn/).

Additional information

Funding

The present study was supported by the National Natural Science Foundation of China (8227151381). The funders had no role in study design, data collection, analysis, publication decision, or manuscript preparation.