Screening,prediction And Optimization Of Anti Breast Cancer Candidate Drugs Based On Sparse Neural Network In High-dimensional Data

Breast cancer is one of the most common malignancies globally,severely impacting women’s health and quality of life.Although there are various chemotherapy drugs available for breast cancer treatment,drug resistance and side effects remain major challenges.Therefore,the successful development of more effective and novel drugs is key to overcoming breast cancer.New drug development typically requires the screening of numerous potential compounds,which is time-consuming and resource-intensive.However,with the rise of big data analysis technology,artificial intelligence is gradually making its mark in assisting drug development,effectively shortening the drug development cycle and cost.This article aims to propose a more comprehensive and scientific new technology to assist in the design,simulation,and optimization of breast cancer candidate drugs.The main research work and achievements are as follows.Firstly,an improved Sparse Input Neural Network（SPINN）is proposed.Traditional neural networks often suffer from the curse of dimensionality when processing high-dimensional data,leading to overfitting and other problems.Therefore,this study improves the SPINN algorithm by adding an adaptive Lasso method to handle high-dimensional data and improve the algorithm’s generalization ability.The sparsity and asymptotic normality are proven theoretically,explaining that the SPINN algorithm can achieve weight convergence to zero in high-dimensional sparse variable selection.A numerical simulation experiment is designed to verify the algorithm’s convergence performance,function approximation ability,regression,and classification ability.All experimental results confirm the effectiveness of the proposed algorithm,providing guidance for algorithm application and enhancing the model’s stability and interpretability.Secondly,a closed-loop drug design assistance scheme is constructed.First,two sets of breast cancer candidate drug screening plans are given from different perspectives.Based on this,the SPINN algorithm is applied to target biological activity prediction and pharmacokinetic behavior prediction model construction.Numerous comparative experiments verify the superiority of the algorithm in this article and deploy the algorithm to predict the p IC₅₀,IC₅₀,and ADMET properties of 50 unknown candidate drugs.Finally,the GWO-KELM algorithm is applied to multi-objective optimization of breast cancer drug molecular discovery,obtaining the optimal feature interval of 30 drug molecules.The research achievements have important practical significance for promoting breast cancer drug development.