Directed evolution is an important research activity in synthetic biology and biotechnology. Numerous reports describe the application of tedious mutation/screening cycles for the improvement of proteins. Recently, knowledge-based approaches have facilitated the prediction of protein properties and the identification of improved mutants. However, epistatic phenomena constitute an obstacle which can impair the predictions in protein engineering. We present an innovative sequence-activity relationship (innov’SAR) methodology based on digital signal processing combining wet-lab experimentation and computational protein design. In our machine learning approach, a predictive model is developed to find the resulting property of the protein when the n single point mutations are permuted (2ⁿ combinations). The originality of our approach is that only sequence information and the fitness of mutants measured in the wet-lab are needed to build models. We illustrate the application of the approach in the case of improving the enantioselectivity of an epoxide hydrolase from Aspergillus niger. n = 9 single point mutants of the enzyme were experimentally assessed for their enantioselectivity and used as a learning dataset to build a model. Based on combinations of the 9 single point mutations (2⁹), the enantioselectivity of these 512 variants were predicted, and candidates were experimentally checked: better mutants with higher enantioselectivity were indeed found.

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. 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. 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.

The selection of optimal enzyme concentration in multienzyme cascade reactions for the highest product yield in practice is very expensive and time-consuming process. The modelling of biological pathways is a difficult process because of the complexity of the system. The mathematical modelling of the system using an analytical approach depends on the many parameters of enzymes which rely on tedious and expensive experiments. The artificial neural network (ANN) method has been successively applied in different fields of science to perform complex functions. In this study, ANN models were trained to predict the flux for the upper part of glycolysis as inferred by NADH consumption, using four enzyme concentrations i.e., phosphoglucoisomerase, phosphofructokinase, fructose-bisphosphate-aldolase, triose-phosphate-isomerase. Out of three ANN algorithms, the neuralnet package with two activation functions, “logistic” and “tanh” were implemented. The prediction of the flux was very efficient: RMSE and R² were 0.847, 0.93 and 0.804, 0.94 respectively for logistic and tanh functions using a cross validation procedure. This study showed that a systemic approach such as ANN could be used for accurate prediction of the flux through the metabolic pathway. This could help to save a lot of time and costs, particularly from an industrial perspective. The R-code is available at: https://github.com/DSIMB/ANN-Glycolysis-Flux-Prediction.

The metabolic engineering of pathways has been used extensively to produce molecules of interest on an industrial scale. Methods like gene regulation or substrate channeling helped to improve the desired product yield. Cell-free systems are used to overcome the weaknesses of engineered strains. One of the challenges in a cell-free system is selecting the optimized enzyme concentration for optimal yield. Here, a machine learning approach is used to select the enzyme concentration for the upper part of glycolysis. The artificial neural network approach (ANN) is known to be inefficient in extrapolating predictions outside the box: high predicted values will bump into a sort of “glass ceiling”. In order to explore this “glass ceiling” space, we developed a new methodology named glass ceiling ANN (GC-ANN). Principal component analysis (PCA) and data classification methods are used to derive a rule for a high flux, and ANN to predict the flux through the pathway using the input data of 121 balances of four enzymes in the upper part of glycolysis. The outcomes of this study are i. in silico selection of optimum enzyme concentrations for a maximum flux through the pathway and ii. experimental in vitro validation of the “out-of-the-box” fluxes predicted using this new approach. Surprisingly, flux improvements of up to 63% were obtained. Gratifyingly, these improvements are coupled with a cost decrease of up to 25% for the assay.

The work aiming to unravel the correlation between protein sequence and function in the absence of structural information can be highly rewarding. We present a new way of considering descriptors from the amino acids index database for modeling and predicting the fitness value of a polypeptide chain. This approach includes the following steps: (i) Calculating Q elementary numerical sequences (Ele_SEQ) depending on the encoding of the amino acid residues, (ii) determining an extended numerical sequence (Ext_SEQ) by concatenating the Q elementary numerical sequences, wherein at least one elementary numerical sequence is a protein spectrum obtained by applying fast Fourier transformation (FFT), and (iii) predicting a value of fitness for polypeptide variants (train and/or validation set). These new descriptors were tested on four sets of proteins of different lengths (GLP-2, TNF alpha, cytochrome P450, and epoxide hydrolase) and activities (cAMP activation, binding affinity, thermostability and enantioselectivity). We show that the use of multiple physicochemical descriptors coupled with the implementation of the FFT, taking into account the interactions between residues of amino acids within the protein sequence, could lead to very significant improvement in the quality of models and predictions. The choice of the descriptor or of the combination of descriptors and/or FFT is dependent on the couple protein/fitness. This approach can provide potential users with value added to existing mutant libraries where screening efforts have so far been unsuccessful in finding improved polypeptide mutants for useful applications.