PyPEF—An Integrated Framework for Data-Driven Protein Engineering

Abstract

Data-driven strategies are gaining increased attention in protein engineering due to recent advances in access to large experimental databanks of proteins, next-generation sequencing (NGS), high-throughput screening (HTS) methods, and the development of artificial intelligence algorithms. However, the reliable prediction of beneficial amino acid substitutions, their combination, and the effect on functional properties remain the most significant challenges in protein engineering, which is applied to develop proteins and enzymes for biocatalysis, biomedicine, and life sciences. Here, we present a general-purpose framework (PyPEF: pythonic protein engineering framework) for performing data-driven protein engineering using machine learning methods combined with techniques from signal processing and statistical physics. PyPEF guides the identification and selection of beneficial proteins of a defined sequence space by systematically or randomly exploring the fitness of variants and by sampling random evolution pathways. The performance of PyPEF was evaluated concerning its predictive accuracy and throughput on four public protein and enzyme data sets using common regression models. It was proved that the program could efficiently predict the fitness of protein sequences for different target properties (predictive models with coefficient of determination values ranging from 0.58 to 0.92). By combining machine learning and protein evolution, PyPEF enabled the screening of proteins with various functions, reaching a screening capacity of more than 500,000 protein sequence variants in the timeframe of only a few minutes on a personal computer. PyPEF displayed significant accuracies on four public data sets (different proteins and properties) and underlined the potential of integrating data-driven technologies for covering different philosophies by either predicting the fitness of the variants to the highest accuracy accounting for epistatic effects or capturing the general trend of introduced mutations on the fitness in directed protein evolution campaigns. In essence, PyPEF can provide a powerful solution to current sequence exploration and combinatorial problems faced in protein engineering through exhaustive in silico screening of the sequence space.

Generalized Property-Based Encoders and Digital Signal Processing Facilitate Predictive Tasks in Protein Engineering

“…Thus, by applying Fourier Transforms, we can capture, to some extent, this spatial dependency. Furthermore, this property results beneficial for any property-based encoding strategy, as previously reported in Siedhoff et al (2021) , Cadet et al (2018b) , and Cosic (1994) . Based on the above, we propose the combination of our encoders together with the application of Fourier transforms in order to improve the performance of predictive models.…”

“…Unsurprisingly, encoders play a fundamental role in the quality of the outcome of predictive models ( Yang et al, 2019 ; Wittmann et al, 2021 ). However, while there is a wide variety of encoding techniques, there is no general agreement on which one to select for a specific task ( Yang et al, 2018 ; Siedhoff et al, 2021 ). The first encoding approaches represented amino acid sequences in discrete manner (numeric-wise), using techniques such as One Hot or Ordinal Encoder ( Winter 1998 ; Pavelka et al, 2009 ; Brownlee, 2020 ).…”

“…Other remarkable examples find applications in cancer studies ( Cosic et al, 2016 ), analysis of conserved motif regions ( Hejase de Trad et al, 2000 ), evaluation of bioactivity ( Cosic, 1994 ), and the prediction of secondary structure and protein-protein interactions ( Cosic et al, 2016 ). Recently, researchers have combined with great success digital signal processing with machine learning to develop predictive models to evaluate—among other variables—enantioselectivity and protein thermostability ( Cadet et al, 2018a , b ; Siedhoff et al, 2021 ).…”

Evolution of E. coli Phytase Toward Improved Hydrolysis of Inositol Tetraphosphate

“…Data-driven recombination of beneficial substitutions, including model training (parameter optimization), validation, and prediction, was performed using the data-driven protein engineering framework PyPEF (Siedhoff et al, 2021). This framework performs sequence-based model training using diverse machine learning algorithms available from the Scikitlearn Python package (Pedregosa et al, 2011).…”

Machine-Learning-Guided Library Design Cycle for Directed Evolution of Enzymes: The Effects of Training Data Composition on Sequence Space Exploration

“…Recently, machine learning (ML) has been used to accelerate directed evolution of proteins. − In this method, saturation mutagenesis and/or random mutagenesis are performed to generate an initial library. The variants in the library are experimentally evaluated to obtain their sequences and functions and then used as training data to construct a ML model that predicts the function from the sequence.…”