PremPLI: a machine learning model for predicting the effects of missense mutations on protein-ligand interactions

Abstract

Resistance to small-molecule drugs is the main cause of the failure of therapeutic drugs in clinical practice. Missense mutations altering the binding of ligands to proteins are one of the critical mechanisms that result in genetic disease and drug resistance. Computational methods have made a lot of progress for predicting binding affinity changes and identifying resistance mutations, but their prediction accuracy and speed are still not satisfied and need to be further improved. To address these issues, we introduce a structure-based machine learning method for quantitatively estimating the effects of single mutations on ligand binding affinity changes (named as PremPLI). A comprehensive comparison of the predictive performance of PremPLI with other available methods on two benchmark datasets confirms that our approach performs robustly and presents similar or even higher predictive accuracy than the approaches relying on first-principle statistical mechanics and mixed physics- and knowledge-based potentials while requires much less computational resources. PremPLI can be used for guiding the design of ligand-binding proteins, identifying and understanding disease driver mutations, and finding potential resistance mutations for different drugs. PremPLI is freely available at https://lilab.jysw.suda.edu.cn/research/PremPLI/ and allows to do large-scale mutational scanning.

Experimental characterization ofin silicored-shift predicted iLOVL470T/Q489Kand iLOVV392K/F410V/A426Smutants

“…We analysed the change of FMN binding affinity to the iLOV protein upon introduction of the mutations used in our study. In line with our observation of lack of FMN binding for constructs bearing the V392K mutation, PremPLI 26 indicates the strongest decrease (1.05 kcal mol -1 ) in the binding affinity towards FMN for mutants with this mutation (Table 2). Taken together, our results suggest that applying simple in silico stability and ligand binding analyses could help determine mutations that would likely lead to fluorescence loss.…”

“…Several studies recently also provided a molecular characterization of the spectral effects of mutations on iLOV to gain a mechanistic understanding [27][28][29] . Röllen et al 29 (and its homolog CagFbFP I52T/Q148K ) was mutated at the same sites as in our study, the V392T mutation does not reduce the affinity of iLOV for FMN as much as V392K does (FMN ΔΔG for V392T 0.67 kcal mol -1 vs for V392K 1.05 kcal mol -1 predicted by PremPLI 26 ). For the rational design of iLOV variants, using in silico predictors could guide researchers to filter out mutations that would disrupt protein stability and/or reduce FMN binding.…”

“…Its homolog mutant CagFbFP I52T/Q148K (emission maximum 504 nm) and another blue-shifted variant CagFbFP Q148K (emission maximum 491 nm) were used in fluorescence microscopy experiments and were successfully spectrally separated. While iLOV V392T/Q489K (and its homolog CagFbFP I52T/Q148K ) was mutated at the same sites as in our study, the V392T mutation does not reduce the affinity of iLOV for FMN as much as V392K does (FMN ΔΔG for V392T 0.67 kcal mol -1 vs for V392K 1.05 kcal mol -1 predicted by PremPLI 26 ). For the rational design of iLOV variants, using in silico predictors could guide researchers to filter out mutations that would disrupt protein stability and/or reduce FMN binding.…”

“…WT iLOV (for amino acid sequence see Supplementary Information) was modeled by the Swiss-Model web-server 21 using PDB id: 4EEP as template. The resulting structure was submitted to DUET 24 and PremPLI 26 web-servers to predict the change of protein stability and FMN binding affinity, respectively.…”

“…Confirming our chaperone co-purification and lack of FMN binding observations, the double (iLOV L470T/Q489K ) and triple (iLOV V392K/F410V/A426S ) mutants were predicted to become unstable upon introduction of the mutations (see Table 2). Additionally, we used PremPLI 26 , a protein structure-based machine learning method for the prediction of the effects of single point mutations on ligand binding affinity. We analysed the change of FMN binding affinity to the iLOV protein upon introduction of the mutations used in our study.…”