Machine Learning Enables Selection of Epistatic Enzyme Mutants for Stability Against Unfolding and Detrimental Aggregation

Li, Guangyue; Qin, Youcai; Fontaine, Nicolas; Chong, Matthieu Ng Fuk; Maria-Solano, Miguel A.; Feixas, Ferran; Cadet, Xavier F.; Pandjaitan, Rudy; Garcia‐Borràs, Marc; Cadet, Frédéric; Reetz, Manfred T.

doi:10.1002/cbic.202000612

Cited by 29 publications

(29 citation statements)

References 74 publications

(109 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…48,72 A structureindependent mutant library screening machine learning approach termed innov'SAR appeared recently. 73 The innov'SAR 37,41,74,75 pipeline applies featurization by Fouriertransforming numerical indices, which represent physicochemical and biochemical properties for each amino acid, taken from the amino acid index (AAindex) database. 76,77 After fast Fourier transform (FFT) processing, a spectral form of the protein is generated and used as the input for subsequent statistical modeling.…”

Section: Introductionmentioning

confidence: 99%

PyPEF—An Integrated Framework for Data-Driven Protein Engineering

Siedhoff

Illig

Schwaneberg

et al. 2021

J. Chem. Inf. Model.

View full text Add to dashboard Cite

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.

show abstract

Section: Introductionmentioning

confidence: 99%

PyPEF—An Integrated Framework for Data-Driven Protein Engineering

Siedhoff

Illig

Schwaneberg

et al. 2021

J. Chem. Inf. Model.

View full text Add to dashboard Cite

show abstract

“…Refinement of computational protein structure prediction using ML+MD has also been reported [heo2020]. With respect to variants that could impact protein function, several studies have used MD and ML to predict drug-resistant variants in Mycobacterium tuberculosis that is involved in tuberculosis [jama2020; mugu2021], variants that promote CTX-M9 mediated antibiotic resistance [lata2017], variants in epistatic enzyme that impact its stability and aggregation [li2021], and impacts of variants on protein binding dynamics [babb2020] or inducing the phenotypic alterations [garg2019].…”

Section: Discussionmentioning

confidence: 99%

Prediction of Kv11.1 potassium channel PAS-domain variants trafficking via machine learning

Immadisetty¹,

Kekenes‐Huskey²

2021

Preprint

View full text Add to dashboard Cite

1AbstractLong QT syndrome (LQTS) is a cardiac condition characterized by a prolonged QT-interval. This prolongation can contribute to fatal arrhythmias despite otherwise normal metrics of cardiac function. Hence, genetic screening of individuals remains a standard for initial identification of individuals susceptible to LQTS. Several genes, including KCNH2 that encodes for the Kv11.1 channel are known to cause LQTS2, however, only a small subset of variants found in the human population are established as pathogenic. Hence, the majority of its missense mutations are known to be benign or are variant of unknown significance (VUS). Here we use molecular dynamics (MD) simulations and machine learning (ML) to determine the propensity for Kv11.1 channel variants to present loss-of-function behavior. Specifically, we use these computational techniques to correlate structural and dynamic changes in an important Kv11.1 subdomain, the PAS-domain (PASD), with the channel’s ability to traffic normally to the plasma membrane. With these techniques, we have identified several molecular features, namely, number of hydrating waters and intra-PASD hydrogen bonds, as moderately predictive of trafficking. Together with bioinformatics data including sequence conservation and folding energies, we are able to predict with reasonable accuracy (≈75%) the ability of VUS’ to traffic. Additionally, we compared two ML algorithms i.e., Decision tree (DT)s and Random forest (RF) for their robustness, and we report that RF performed moderately better than DTs’. Features derived from MD trajectories particularly help improve the prediction of trafficking deficient variants by both ML techniques.

show abstract

“…The generic pipeline of a protein engineering campaign supported by machine learning consists of the generation of experimental data representing sequence-phenotype pairs, and training a statistical or machine learning model to predict phenotypes from input sequences never assayed before. The phenotypic property of interest can be chosen from several features, including catalytic properties, 98,99 substrate affinity, stability, 100–102 and expression level in the host organism. Prominent examples of machine learning assisted protein engineering include membrane channel engineering, 103,104 protein structure prediction 105,106 and protein–protein interactions.…”

Section: Machine Learning In Enzyme Engineeringmentioning

confidence: 99%