Soft threshold partial least squares predicts the survival fraction of malignant glioma cells against different concentrations of methotrexate’s derivatives

Abstract

Abstract Chemotherapy appeared to be a significant advancement in cancer research, with fewer side effects. Methotrexate (MTX) is a widely used anticancer drug with strong activity but serious side effects. Several MTX derivatives have been reported, with modifications at various sites to reduce side effects and increase efficacy. The current study uses FTIR spectroscopy to predict the survival fraction of human malignant glioma U87 (MG-U87) cell lines against MTX derivatives. Together with Parent MTX several aldehydes viz. Benzaldehyde, Chlorobenzaldehyde, 2-Chlorobenzaldehyde, 3-Nitrobenzaldehyde, 5-Chloro-2-hydroxybenz-aldehyde, 2-Hydroxy-5-Nitrobenzaldehyde, 2-Thiocarboxyaldehyde, Trans-2-pentenal, and Glutaraldehyde are treated with MTX to obtain MTX derivatives. The prediction of survival fraction of malignant glioma cells is carried out by Lasso, Elastic net and Soft PLS at different concentration levels of synthesized derivatives, including 400 μM, 200 μM, 100 μM, 50 μM, 25 μM and 12.5 μM. The cross-validated prediction error is minimised to optimise spectral wavelength selection and model parameters. It appears that the RMSE computed from test data is significantly varying with the change of models (p = 0.012), with the change of concentrations levels (p $$\le 0.001$$ ≤ 0.001 ) and with the change of combination of models and concentration level (p $$\le 0.001$$ ≤ 0.001 ). StPLS outperforms in predicting survival fraction of glioma cells at the concentration level 50 μM, 100 μM and 400 μM respectively with relative RMSE = 0.1,0.14 and 0.55. Lasso outperforms at the concentration level 12.5 μM, and 200 μM respectively with relative RMSE = 0.4 and 0.14. Elastic net outperforms at the concentration level 25 μM with relative RMSE = 0.8. Consistently appeared influential wavelength identifies the influential functional compounds which best predicts the survival fraction. Hence FTIR appears potential candidate for estimating survival fraction of MTX derivatives.