Predicting Anticancer Drug Resistance Mediated by Mutations

Abstract

Cancer drug resistance presents a challenge for precision medicine. Drug-resistant mutations are always emerging. In this study, we explored the relationship between drug-resistant mutations and drug resistance from the perspective of protein structure. By combining data from previously identified drug-resistant mutations and information of protein structure and function, we used machine learning-based methods to build models to predict cancer drug resistance mutations. The performance of our combined model achieved an accuracy of 86%, a Matthews correlation coefficient score of 0.57, and an F1 score of 0.66. We have constructed a fast, reliable method that predicts and investigates cancer drug resistance in a protein structure. Nonetheless, more information is needed concerning drug resistance and, in particular, clarification is needed about the relationships between the drug and the drug resistance mutations in proteins. Highly accurate predictions regarding drug resistance mutations can be helpful for developing new strategies with personalized cancer treatments. Our novel concept, which combines protein structure information, has the potential to elucidate physiological mechanisms of cancer drug resistance.

Keywords: cancer drug; drug resistance; feature selection; machine learning; personalized therapeutics; protein structure; single amino acid variation.

Figure 1

Simulated and crystal structures of the BRAF-drug complex. (a) The simulated structure of the BRAF-vemurafenib complex. Docked vemurafenib is indicated by the orange-colored stick. The magenta-colored stick represents the pyrazolopyridine inhibitor, which is located in the crystal structure of the BRAF-pyrazolopyridine inhibitor complex. The gray-colored cartoon structures of BRAF (PDBID: 3TV6 [65]) were drawn using PyMOL software. Residues found within 8 Å from vemurafenib are represented by blue coloring. The drug-resistant SAV (L505) is represented as spheres in blue. (b) Simulation of the amino acid mutated to histidine (H505) is shown in the color green.

Figure 2

Simulated and crystal structures of the MAP2K2-drug complex. (a) The simulated structure of the MAP2K2-PD0325901 complex. Docked PD0325901 is indicated by the yellow-colored stick. The blue-colored stick indicates the PD184352-like inhibitor. The gray-colored cartoon structures of MAP2K2 (PDBID: 1S9I [66]) were drawn using PyMOL software. Residues found within 8 Å from PD0325901 are represented by pink coloring. The drug-resistant SAV (V215) is represented as spheres in pink. (b) Simulation of mutated glutamic acid (E215) is shown in the color green.

Figure 3

Simulated and crystal structures of the ROS1-drug complex. (a) The simulated structure of the ROS1-crizotinib structure complex. Docked crizotinib is indicated by the yellow-colored stick. The transparent orange-colored stick indicates crizotinib. The gray-colored cartoon structures of ROS1 (PDBID: 3ZBF [67]) were drawn using PyMOL software. Residues found within 8 Å from crizotinib are represented by green coloring. The drug-resistant SAV (G2032) is represented as spheres. (b) Simulation of mutated arginine (R2032) is shown in the color purple.

Figure 4

Selected features that were applied in the four drug prediction models.

Figure 5

Simulated structures of the protein-drug complexes with mutation alterations. Wild-type amino acids are represented as spheres and mutated-type amino acids as sticks. (a) The BRAF-vemurafenib complex with the F516L, M517V, and G596S mutations. (b) The MAP2K2-PD0325901 structure complex with the G83S mutation. (c) The ROS1-crizotinib structure complex with the S2088F mutation.

Figure 6

Distributions of drug-resistant and non-drug-resistant SAVs in the training set.

Figure 7

The workflow diagram represents the study’s prediction system for cancer drug resistance.