厦门大学学报（自然科学版）2025，Vol.64Issue(5)：886-896,11.DOI:10.6043/j.issn.0438-0479.202404001
机器学习指导下的椎间盘组织工程体系预测模型建立
Construction of a prediction model of intervertebral disc tissue engineering system guided by machine learning
摘要
Abstract
[Objective]Intervertebral discs degeneration(IVDD)is a major contributor to back,neck,and radicular pain,which mainly involves the pathological process of dehydration and degeneration of the nucleus pulposus,weak rupture of the annulus fibrosus,and calcification of the cartilage endplate defect.These degenerative changes lead to compromised structural integrity and function of the intervertebral discs,resulting in significant pain and disability.Current treatment options for IVDD are limited and often focus on symptom management rather than addressing the underlying pathology.Intervertebral disc tissue engineering(IDTE)has emerged as a promising approach to improve outcomes of IVDD by promoting the regeneration of disc tissues and restoring their function.This study aims to construct a prediction model of the biological effects of intervertebral disc scaffolds using various machine learning(ML)methods.The goal is to reduce experimental costs associated with trial and error,and to achieve efficient treatment of tissue engineering for intervertebral disc degenerative diseases.[Methods]To develop a robust prediction model,data were mined from literature on intervertebral disc scaffold-related materials,technological processes,and cell culture conditions published over the past decade.This comprehensive data set included multiple features and diverse labels,capturing a wide range of variables that influence scaffold performance and cell behavior.The data set was then used to train and evaluate the performance of 21 classification algorithms,including ridge classifier,Gaussian naive Bayes(Gaussian NB),linear discriminant analysis(LDA),and Gaussian process classifier(GPC).Each algorithm was assessed for predictive accuracy,generalization ability,and computational efficiency.[Results]Through literature retrieval,data extracting,and data preprocessing,a data set containing 10 principle scaffold materials,5 modeling ingredients,5 bioactive factors,5 scaffold forms,and 10 cell types was built.Among all features,gelatin,tetrazine-norbornene(Tz-Nb),transforming growth factor β(TGF-β),hydrogel and human nucleus pulposus cells were most frequently applied.As for targets of the data set,69 cases were matched with label 0,suggesting unsuccessful modeling of IDTE scaffolds under specific feature combination.60 cases with label 1 were identified due to the insignificant improvement of biological effect for IVDD-related cells compared to conventional cultural environment.Finally,82 cases of scaffold systems were proved to have positive effects on cell behavior,which were labeled with 2.After data input,ML training and predicting,a series of models indicating the cellular effects of IDTE scaffolds were established based on the data set.The results demonstrated that most of the included classification models showed relatively consistent performances,with relatively short execution time.Among them,the ML model based on ridge classifier algorithm provided the best accurate prediction for cell behavior on scaffolds.This model was found to have a high generalization value,indicating its potential for widespread application in IDTE.In addition,the shapley additive explanations(SHAP)model was applied to attribute each IDTE features to bio-target labels.[Conclusion]In summary,ML methods hold significant promise for guiding the design and exploitation of IDTE systems.Through appropriate model construction,researchers can selectively combine features to design scaffolds with good biological activity quickly and easily.The use of prediction model can streamline the development process,reduce the reliance on extensive empirical testing,and accelerate the translation of IDTE from the laboratory to clinical applications.By leveraging the predictive power of ML,it is possible to identify optimal scaffold designs that promote tissue regeneration and improve patient outcomes.This approach represents a major advancement in the field of tissue engineering and may offer a new avenue for the effective treatment of IVDD.关键词
椎间盘退行性病变/机器学习/组织工程/支架Key words
intervertebral disc degeneration/machine learning/tissue engineering/scaffold分类
医药卫生引用本文复制引用
韩照普,尹正,叶晓健..机器学习指导下的椎间盘组织工程体系预测模型建立[J].厦门大学学报（自然科学版）,2025,64(5):886-896,11.基金项目
上海交通大学医工交叉重点项目(YG2021ZD34) （YG2021ZD34）
