Computational analyses of drug resistance mutations in katG and emb complexes in Mycobacterium tuberculosis.

The number of antibiotic resistant pathogens is increasing rapidly, and with this comes a substantial socioeconomic cost that threatens much of the world. To alleviate this problem, we must use antibiotics in a more responsible and informed way, further our understanding of the molecular basis of drug resistance, and design new antibiotics. Here, we focus on a key drug-resistant pathogen, Mycobacterium tuberculosis, and computationally analyze trends in drug-resistant mutations in genes of the proteins embA, embB, embC, and katG, which play essential roles in the action of the first-line drugs ethambutol and isoniazid. We use docking to predict binding modes of isoniazid to katG that agree with suggested binding sites found in our laboratory using cryo-EM. Using mutant stability predictions, we recapitulate the idea that resistance occurs when katG's heme cofactor is destabilized rather than due to a decrease in affinity to isoniazid. Conversely, we have identified resistance mutations that affect the affinity of ethambutol more drastically than the affinity of the natural substrate of embB. With this, we illustrate that we can distinguish between the two types of drug resistance-cofactor destabilization and drug affinity reduction-suggesting potential uses in the prediction of novel drug-resistant mutations.

DeepSTABp: A Deep Learning Approach for the Prediction of Thermal Protein Stability.

Proteins are essential macromolecules that carry out a plethora of biological functions. The thermal stability of proteins is an important property that affects their function and determines their suitability for various applications. However, current experimental approaches, primarily thermal proteome profiling, are expensive, labor-intensive, and have limited proteome and species coverage. To close the gap between available experimental data and sequence information, a novel protein thermal stability predictor called DeepSTABp has been developed. DeepSTABp uses a transformer-based protein language model for sequence embedding and state-of-the-art feature extraction in combination with other deep learning techniques for end-to-end protein melting temperature prediction. DeepSTABp can predict the thermal stability of a wide range of proteins, making it a powerful and efficient tool for large-scale prediction. The model captures the structural and biological properties that impact protein stability, and it allows for the identification of the structural features that contribute to protein stability. DeepSTABp is available to the public via a user-friendly web interface, making it accessible to researchers in various fields.

Protein stability engineering insights revealed by domain-wide comprehensive mutagenesis.

The accurate prediction of protein stability upon sequence mutation is an important but unsolved challenge in protein engineering. Large mutational datasets are required to train computational predictors, but traditional methods for collecting stability data are either low-throughput or measure protein stability indirectly. Here, we develop an automated method to generate thermodynamic stability data for nearly every single mutant in a small 56-residue protein. Analysis reveals that most single mutants have a neutral effect on stability, mutational sensitivity is largely governed by residue burial, and unexpectedly, hydrophobics are the best tolerated amino acid type. Correlating the output of various stability-prediction algorithms against our data shows that nearly all perform better on boundary and surface positions than for those in the core and are better at predicting large-to-small mutations than small-to-large ones. We show that the most stable variants in the single-mutant landscape are better identified using combinations of 2 prediction algorithms and including more algorithms can provide diminishing returns. In most cases, poor in silico predictions were tied to compositional differences between the data being analyzed and the datasets used to train the algorithm. Finally, we find that strategies to extract stabilities from high-throughput fitness data such as deep mutational scanning are promising and that data produced by these methods may be applicable toward training future stability-prediction tools.

PON-tstab: Protein Variant Stability Predictor. Importance of Training Data Quality.

Several methods have been developed to predict effects of amino acid substitutions on protein stability. Benchmark datasets are essential for method training and testing and have numerous requirements including that the data is representative for the investigated phenomenon. Available machine learning algorithms for variant stability have all been trained with ProTherm data. We noticed a number of issues with the contents, quality and relevance of the database. There were errors, but also features that had not been clearly communicated. Consequently, all machine learning variant stability predictors have been trained on biased and incorrect data. We obtained a corrected dataset and trained a random forests-based tool, PON-tstab, applicable to variants in any organism. Our results highlight the importance of the benchmark quality, suitability and appropriateness. Predictions are provided for three categories: stability decreasing, increasing and those not affecting stability.

Protein Stability Perturbation Contributes to the Loss of Function in Haploinsufficient Genes.

Missense variants are among the most studied genome modifications as disease biomarkers. It has been shown that the "perturbation" of the protein stability upon a missense variant (in terms of absolute ΔΔG value, i.e., |ΔΔG|) has a significant, but not predictive, correlation with the pathogenicity of that variant. However, here we show that this correlation becomes significantly amplified in haploinsufficient genes. Moreover, the enrichment of pathogenic variants increases at the increasing protein stability perturbation value. These findings suggest that protein stability perturbation might be considered as a potential cofactor in diseases associated with haploinsufficient genes reporting missense variants.

Computational Modeling of Protein Stability: Quantitative Analysis Reveals Solutions to Pervasive Problems.

Accurate modeling of the effects of mutations on protein stability is central to understanding and controlling proteins in myriad natural and applied contexts. Here, we reveal through rigorous quantitative analysis that stability prediction tools often favor mutations that increase stability at the expense of solubility. Moreover, while these tools may accurately identify strongly destabilizing mutations, the experimental effect of mutations predicted to stabilize is actually near neutral on average. The commonly used "classification accuracy" metric obscures this reality; accordingly, we recommend performance measures, such as the Matthews correlation coefficient (MCC). We demonstrate that an absurdly simple machine-learning algorithm-a neural network of just two neurons-unexpectedly achieves high classification accuracy, but its inadequacies are revealed by a low MCC. Despite the above limitations, making multiple mutations markedly improves the prospects for achieving a stabilization target, and modest improvements in the precision of future tools may yield disproportionate gains.

Multistate Computational Protein Design with Backbone Ensembles.

The ability of computational protein design (CPD) to identify protein sequences possessing desired characteristics in vast sequence spaces makes it a highly valuable tool in the protein engineering toolbox. CPD calculations are typically performed using a single-state design (SSD) approach in which amino-acid sequences are optimized on a single protein structure. Although SSD has been successfully applied to the design of numerous protein functions and folds, the approach can lead to the incorrect rejection of desirable sequences because of the combined use of a fixed protein backbone template and a set of rigid rotamers. This fixed backbone approximation can be addressed by using multistate design (MSD) with backbone ensembles. MSD improves the quality of predicted sequences by using ensembles approximating conformational flexibility as input templates instead of a single fixed protein structure. In this chapter, we present a step-by-step guide to the implementation and analysis of MSD calculations with backbone ensembles. Specifically, we describe ensemble generation with the PertMin protocol, execution of MSD calculations for recapitulation of Streptococcal protein G domain ß1 mutant stability, and analysis of computational predictions by sequence binning. Furthermore, we provide a comparison between MSD and SSD calculation results and discuss the benefits of multistate approaches to CPD.

Optimization of rotamers prior to template minimization improves stability predictions made by computational protein design.

Computational protein design (CPD) predictions are highly dependent on the structure of the input template used. However, it is unclear how small differences in template geometry translate to large differences in stability prediction accuracy. Herein, we explored how structural changes to the input template affect the outcome of stability predictions by CPD. To do this, we prepared alternate templates by Rotamer Optimization followed by energy Minimization (ROM) and used them to recapitulate the stability of 84 protein G domain ß1 mutant sequences. In the ROM process, side-chain rotamers for wild-type (WT) or mutant sequences are optimized on crystal or nuclear magnetic resonance (NMR) structures prior to template minimization, resulting in alternate structures termed ROM templates. We show that use of ROM templates prepared from sequences known to be stable results predominantly in improved prediction accuracy compared to using the minimized crystal or NMR structures. Conversely, ROM templates prepared from sequences that are less stable than the WT reduce prediction accuracy by increasing the number of false positives. These observed changes in prediction outcomes are attributed to differences in side-chain contacts made by rotamers in ROM templates. Finally, we show that ROM templates prepared from sequences that are unfolded or that adopt a nonnative fold result in the selective enrichment of sequences that are also unfolded or that adopt a nonnative fold, respectively. Our results demonstrate the existence of a rotamer bias caused by the input template that can be harnessed to skew predictions toward sequences displaying desired characteristics.