3D proteochemometrics: using three-dimensional information of proteins and ligands to address aspects of the selectivity of serine proteases

How to approach machine learning-based prediction of drug/compound-target interactions

The identification of drug/compound-target interactions (DTIs) constitutes the basis of drug discovery, for which computational predictive approaches have been developed. As a relatively new data-driven paradigm, proteochemometric (PCM) modeling utilizes both protein and compound properties as a pair at the input level and processes them via statistical/machine learning. The representation of input samples (i.e., proteins and their ligands) in the form of quantitative feature vectors is crucial for the extraction of interaction-related properties during the artificial learning and subsequent prediction of DTIs. Lately, the representation learning approach, in which input samples are automatically featurized via training and applying a machine/deep learning model, has been utilized in biomedical sciences. In this study, we performed a comprehensive investigation of different computational approaches/techniques for protein featurization (including both conventional approaches and the novel learned embeddings), data preparation and exploration, machine learning-based modeling, and performance evaluation with the aim of achieving better data representations and more successful learning in DTI prediction. For this, we first constructed realistic and challenging benchmark datasets on small, medium, and large scales to be used as reliable gold standards for specific DTI modeling tasks. We developed and applied a network analysis-based splitting strategy to divide datasets into structurally different training and test folds. Using these datasets together with various featurization methods, we trained and tested DTI prediction models and evaluated their performance from different angles. Our main findings can be summarized under 3 items: (i) random splitting of datasets into train and test folds leads to near-complete data memorization and produce highly over-optimistic results, as a result, should be avoided, (ii) learned protein sequence embeddings work well in DTI prediction and offer high potential, despite interaction-related properties (e.g., structures) of proteins are unused during their self-supervised model training, and (iii) during the learning process, PCM models tend to rely heavily on compound features while partially ignoring protein features, primarily due to the inherent bias in DTI data, indicating the requirement for new and unbiased datasets. We hope this study will aid researchers in designing robust and high-performing data-driven DTI prediction systems that have real-world translational value in drug discovery.

Proteochemometrics - recent developments in bioactivity and selectivity modeling

Proteochemometrics is a machine learning based modeling approach relying on a combination of ligand and protein descriptors. With ongoing developments in machine learning and increases in public data the technique is more frequently applied in early drug discovery, typically in ligand-target binding prediction. Common applications include improvements to single target quantitative structure-activity relationship models, protein selectivity and promiscuity modeling, and large-scale deep learning approaches. The increase in predictive power using proteochemometrics is observed in multi-target bioactivity modeling, opening the door to more extensive studies covering whole protein families. On top of that, with deep learning fueling more complex and larger scale models, proteochemometrics allows faster and higher quality computational models supporting the design, make, test cycle.

Application of fourier transform and proteochemometrics principles to protein engineering

Background

Connecting the dots between the protein sequence and its function is of fundamental interest for protein engineers. In-silico methods are useful in this quest especially when structural information is not available. In this study we propose a mutant library screening tool called iSAR (innovative Sequence Activity Relationship) that relies on the physicochemical properties of the amino acids, digital signal processing and partial least squares regression to uncover these sequence-function correlations.

Results

We show that the digitalized representation of the protein sequence in the form of a Fourier spectrum can be used as an efficient descriptor to model the sequence-activity relationship of proteins. The iSAR methodology that we have developed identifies high fitness mutants from mutant libraries relying on physicochemical properties of the amino acids, digital signal processing and regression techniques. iSAR correlates variations caused by mutations in spectra with biological activity/fitness. It takes into account the impact of mutations on the whole spectrum and does not focus on local fitness alone. The utility of the method is illustrated on 4 datasets: cytochrome P450 for thermostability, TNF-alpha for binding affinity, GLP-2 for potency and enterotoxins for thermostability. The choice of the datasets has been made such as to illustrate the ability of the method to perform when limited training data is available and also when novel mutations appear in the test set, that have not been featured in the training set.

Conclusion

The combination of Fast Fourier Transform and Partial Least Squares regression is efficient in capturing the effects of mutations on the function of the protein. iSAR is a fast algorithm which can be implemented with limited computational resources and can make effective predictions even if the training set is limited in size.

Electronic supplementary material

The online version of this article (10.1186/s12859-018-2407-8) contains supplementary material, which is available to authorized users.