RBF神经网络与癌细胞黏附度活性检测研究

Assessment of Cancer Cell Viability Based on RBF Network and Adhesion Strength

目的:研究癌细胞黏附强度与其活性的相关性,并利用机器学习算法提高细胞活性检测方法的准确性。方法:从同一批癌细胞培养皿中分离出两个样本组,一个样本组利用设计制作的离心式微流控芯片,建立癌细胞黏附脱离动态曲线,并提取细胞多黏附强度信息(τ20、τ50、τ80);另一个样本组细胞进行药物反应实验,得到药物反应的半抑制浓度(IC_(50)),利用RBF神经网络算法,将癌细胞多黏附信息作为输入,癌细胞半抑制浓度IC_(50)作为输出,建立细胞活性评估预测模型。结果:相比于传统细胞计数法,基于多黏附强度信息融合的细胞活性评估方法提高了17.2%,该方法具有统计学意义(P<0.05)。结论:利用细胞多黏附特征融合评估细胞活性的方法,有助于提高细胞活性检测精度,对抗癌新药物测试、细胞毒理学实验以及其他生化反应刺激实验具有至关重要的作用。

Objective:To study the correlation between adhesion strength and viability of cancer cells,and use machine learning algorithm to improve the accuracy of cell activity detection method.Methods:Two sample groups were separated from the same batch of cancer cell culture dishes.In one sample group,the dynamic curve of cancer cell adhesion and detachment was established by using the designed centrifugal microfluidic chip,and the information of cell multi-adhesion strength was extracted(τ20,τ50,τ80).Another sample group of cells was carried out drug reaction experiment to obtain the half inhibitory concentration(IC_(50))of drug reaction.Using RBF neural network algorithm,the multi-adhesion information of cancer cells was taken as the input and the IC_(50) of cancer cells was taken as the output to establish the cell viability evaluation and prediction model.Results:Compared with the traditional cell counting method,the cell viability evaluation method based on multi-adhesion strength information fusion increased by 17.2%(P<0.05).Conclusion:The method of evaluating cell viability by using the fusion of cell multi adhesion characteristics is helpful to improve the detection accuracy of cell activity.It plays an important role in the test of new anti-cancer drugs,cell toxicology and other biochemical reaction stimulation experiments.