Novel Strategy for Protein Exploration: High-throughput Screening Assisted with Fuzzy Neural Network

Deep learning in structural bioinformatics: current applications and future perspectives

In this review article, we explore the transformative impact of deep learning (DL) on structural bioinformatics, emphasizing its pivotal role in a scientific revolution driven by extensive data, accessible toolkits and robust computing resources. As big data continue to advance, DL is poised to become an integral component in healthcare and biology, revolutionizing analytical processes. Our comprehensive review provides detailed insights into DL, featuring specific demonstrations of its notable applications in bioinformatics. We address challenges tailored for DL, spotlight recent successes in structural bioinformatics and present a clear exposition of DL-from basic shallow neural networks to advanced models such as convolution, recurrent, artificial and transformer neural networks. This paper discusses the emerging use of DL for understanding biomolecular structures, anticipating ongoing developments and applications in the realm of structural bioinformatics.

Artificial Intelligence Aided Lipase Production and Engineering for Enzymatic Performance Improvement

With the development of artificial intelligence (AI), tailoring methods for enzyme engineering have been widely expanded. Additional protocols based on optimized network models have been used to predict and optimize lipase production as well as properties, namely, catalytic activity, stability, and substrate specificity. Here, different network models and algorithms for the prediction and reforming of lipase, focusing on its modification methods and cases based on AI, are reviewed in terms of both their advantages and disadvantages. Different neural networks coupled with various algorithms are usually applied to predict the maximum yield of lipase by optimizing the external cultivations for lipase production, while one part is used to predict the molecule variations affecting the properties of lipase. However, few studies have directly utilized AI to engineer lipase by affecting the structure of the enzyme, and a set of research gaps needs to be explored. Additionally, future perspectives of AI application in enzymes, including lipase engineering, are deduced to help the redesign of enzymes and the reform of new functional biocatalysts. This review provides a new horizon for developing effective and innovative AI tools for lipase production and engineering and facilitating lipase applications in the food industry and biomass conversion.

Enhancement of protein thermostability by three consecutive mutations using loop-walking method and machine learning

We developed a method to improve protein thermostability, “loop-walking method”. Three consecutive positions in 12 loops of Burkholderia cepacia lipase were subjected to random mutagenesis to make 12 libraries. Screening allowed us to identify L7 as a hot-spot loop having an impact on thermostability, and the P233G/L234E/V235M mutant was found from 214 variants in the L7 library. Although a more excellent mutant might be discovered by screening all the 8000 P233X/L234X/V235X mutants, it was difficult to assay all of them. We therefore employed machine learning. Using thermostability data of the 214 mutants, a computational discrimination model was constructed to predict thermostability potentials. Among 7786 combinations ranked in silico, 20 promising candidates were selected and assayed. The P233D/L234P/V235S mutant retained 66% activity after heat treatment at 60 °C for 30 min, which was higher than those of the wild-type enzyme (5%) and the P233G/L234E/V235M mutant (35%).

Directed Evolution of a Selective and Sensitive Serotonin Sensor via Machine Learning

Serotonin plays a central role in cognition and is the target of most pharmaceuticals for psychiatric disorders. Existing drugs have limited efficacy; creation of improved versions will require better understanding of serotonergic circuitry, which has been hampered by our inability to monitor serotonin release and transport with high spatial and temporal resolution. We developed and applied a binding-pocket redesign strategy, guided by machine learning, to create a high-performance, soluble, fluorescent serotonin sensor (iSeroSnFR), enabling optical detection of millisecond-scale serotonin transients. We demonstrate that iSeroSnFR can be used to detect serotonin release in freely behaving mice during fear conditioning, social interaction, and sleep/wake transitions. We also developed a robust assay of serotonin transporter function and modulation by drugs. We expect that both machine-learning-guided binding-pocket redesign and iSeroSnFR will have broad utility for the development of other sensors and in vitro and in vivo serotonin detection, respectively.

Combinational risk factors of metabolic syndrome identified by fuzzy neural network analysis of health-check data

Background

Lifestyle-related diseases represented by metabolic syndrome develop as results of complex interaction. By using health check-up data from two large studies collected during a long-term follow-up, we searched for risk factors associated with the development of metabolic syndrome.

Methods

In our original study, we selected 77 case subjects who developed metabolic syndrome during the follow-up and 152 healthy control subjects who were free of lifestyle-related risk components from among 1803 Japanese male employees. In a replication study, we selected 2196 case subjects and 2196 healthy control subjects from among 31343 other Japanese male employees. By means of a bioinformatics approach using a fuzzy neural network (FNN), we searched any significant combinations that are associated with MetS. To ensure that the risk combination selected by FNN analysis was statistically reliable, we performed logistic regression analysis including adjustment.

Results

We selected a combination of an elevated level of γ-glutamyltranspeptidase (γ-GTP) and an elevated white blood cell (WBC) count as the most significant combination of risk factors for the development of metabolic syndrome. The FNN also identified the same tendency in a replication study. The clinical characteristics of γ-GTP level and WBC count were statistically significant even after adjustment, confirming that the results obtained from the fuzzy neural network are reasonable. Correlation ratio showed that an elevated level of γ-GTP is associated with habitual drinking of alcohol and a high WBC count is associated with habitual smoking.

Conclusions

This result obtained by fuzzy neural network analysis of health check-up data from large long-term studies can be useful in providing a personalized novel diagnostic and therapeutic method involving the γ-GTP level and the WBC count.

Structure-based drug design to augment hit discovery

Several technology-based strategies have been developed to address the significance of the two phases of drug discovery: hit identification and lead identification. Structure-based drug design (SBDD), a method that depends on possessing the knowledge of 3D structures of biological targets, is growing swiftly with the development of new technologies for searching potential ways to combat disease. The past decade has evidenced a threefold increase in the amount of software and tools in the online repositories. Herein, we review the in silico strategies and modules applied at the level of hit identification and confer the different challenges with possible solutions in enhancing the success rate of the 'hit-to-lead' phase that could eventually help the progress of SBDD in the drug discovery arena.

Mutagenesis Objective Search and Selection Tool (MOSST): an algorithm to predict structure-function related mutations in proteins

Background

Functionally relevant artificial or natural mutations are difficult to assess or predict if no structure-function information is available for a protein. This is especially important to correctly identify functionally significant non-synonymous single nucleotide polymorphisms (nsSNPs) or to design a site-directed mutagenesis strategy for a target protein. A new and powerful methodology is proposed to guide these two decision strategies, based only on conservation rules of physicochemical properties of amino acids extracted from a multiple alignment of a protein family where the target protein belongs, with no need of explicit structure-function relationships.

Results

A statistical analysis is performed over each amino acid position in the multiple protein alignment, based on different amino acid physical or chemical characteristics, including hydrophobicity, side-chain volume, charge and protein conformational parameters. The variances of each of these properties at each position are combined to obtain a global statistical indicator of the conservation degree of each property. Different types of physicochemical conservation are defined to characterize relevant and irrelevant positions. The differences between statistical variances are taken together as the basis of hypothesis tests at each position to search for functionally significant mutable sites and to identify specific mutagenesis targets. The outcome is used to statistically predict physicochemical consensus sequences based on different properties and to calculate the amino acid propensities at each position in a given protein. Hence, amino acid positions are identified that are putatively responsible for function, specificity, stability or binding interactions in a family of proteins. Once these key functional positions are identified, position-specific statistical distributions are applied to divide the 20 common protein amino acids in each position of the protein's primary sequence into a group of functionally non-disruptive amino acids and a second group of functionally deleterious amino acids.

Conclusions

With this approach, not only conserved amino acid positions in a protein family can be labeled as functionally relevant, but also non-conserved amino acid positions can be identified to have a physicochemically meaningful functional effect. These results become a discriminative tool in the selection and elaboration of rational mutagenesis strategies for the protein. They can also be used to predict if a given nsSNP, identified, for instance, in a genomic-scale analysis, can have a functional implication for a particular protein and which nsSNPs are most likely to be functionally silent for a protein. This analytical tool could be used to rapidly and automatically discard any irrelevant nsSNP and guide the research focus toward functionally significant mutations. Based on preliminary results and applications, this technique shows promising performance as a valuable bioinformatics tool to aid in the development of new protein variants and in the understanding of function-structure relationships in proteins.

Screening of peptides with a high affinity for ZnO using spot-synthesized peptide arrays and computational analysis

Metal-binding peptides have attracted attention for their usefulness in biomineralization processes. In this work screening of high affinity peptides against ZnO was performed using a combinatorial library approach with a peptide array and computational analysis. The computationally assisted peptide screening and design enabled identification of linear peptides with affinity for ZnO with limited experimentation. Starting with the screening data obtained from a random 6-mer library of 420 sequences, the characteristics of the peptides with high ZnO affinity were analyzed using a fuzzy neural network algorithm by comparison of high and low affinity peptides. Three physical properties of amino acids (hydrophobicity, isoelectric point and size) and positional information of each residue were analyzed and a peptide rule with restricted amino acids at certain positions was extracted. The average affinity for ZnO increased 2.0-fold when 300 sequences were synthesized according to the restricted random library, compared with the non-restricted library. In addition, for a peptide library with the amino acids at the restricted sites exchanged the average binding capacity decreased to 0.7-fold. Interestingly, the peptides with high ZnO affinity obtained exhibited binding specificity for ZnO. Computationally assisted screening is an invaluable means for finding peptides with limited experimentation.

Control of lipase enantioselectivity by engineering the substrate binding site and access channel

Lipase from Burkholderia cepacia (BCL) has proven to be a very useful biocatalyst for the resolution of 2-substituted racemic acid derivatives, which are important chiral building blocks. Our previous work showed that enantioselectivity of the wild-type BCL could be improved by chemical engineering of the substrate's molecular structure. From this earlier study, three amino acids (L17, V266 and L287) were proposed as targets for mutagenesis aimed at tailoring enzyme enantioselectivity. In the present work, a small library of 57 BCL single mutants targeted on these three residues was constructed and screened for enantioselectivity towards (R,S)-2-chloro ethyl 2-bromophenylacetate. This led to the fast isolation of three single mutants with a remarkable tenfold enhanced or reversed enantioselectivity. Analysis of substrate docking and access trajectories in the active site was then performed. From this analysis, the construction of 13 double mutants was proposed. Among them, an outstanding improved mutant of BCL was isolated that showed an E value of 178 and a 15-fold enhanced specific activity compared to the parental enzyme; thus, this study demonstrates the efficiency of the semirational engineering strategy.

Prediction of protein function improving sequence remote alignment search by a fuzzy logic algorithm

The functional annotation of the new protein sequences represents a major drawback for genomic science. The best way to suggest the function of a protein from its sequence is by finding a related one for which biological information is available. Current alignment algorithms display a list of protein sequence stretches presenting significant similarity to different protein targets, ordered by their respective mathematical scores. However, statistical and biological significance do not always coincide, therefore, the rearrangement of the program output according to more biological characteristics than the mathematical scoring would help functional annotation. A new method that predicts the putative function for the protein integrating the results from the PSI-BLAST program and a fuzzy logic algorithm is described. Several protein sequence characteristics have been checked in their ability to rearrange a PSI-BLAST profile according more to their biological functions. Four of them: amino acid content, matched segment length and hydropathic and flexibility profiles positively contributed, upon being integrated by a fuzzy logic algorithm into a program, BYPASS, to the accurate prediction of the function of a protein from its sequence.

Computationally assisted screening and design of cell-interactive peptides by a cell-based assay using peptide arrays and a fuzzy neural network algorithm

We developed a method of effective peptide screening that combines experiments and computational analysis. The method is based on the concept that screening efficiency can be enhanced from even limited data by use of a model derived from computational analysis that serves as a guide to screening and combining the model with subsequent repeated experiments. Here we focus on cell-adhesion peptides as a model application of this peptide-screening strategy. Cell-adhesion peptides were screened by use of a cell-based assay of a peptide array. Starting with the screening data obtained from a limited, random 5-mer library (643 sequences), a rule regarding structural characteristics of cell-adhesion peptides was extracted by fuzzy neural network (FNN) analysis. According to this rule, peptides with unfavored residues in certain positions that led to inefficient binding were eliminated from the random sequences. In the restricted, second random library (273 sequences), the yield of cell-adhesion peptides having an adhesion rate more than 1.5-fold to that of the basal array support was significantly high (31%) compared with the unrestricted random library (20%). In the restricted third library (50 sequences), the yield of cell-adhesion peptides increased to 84%. We conclude that a repeated cycle of experiments screening limited numbers of peptides can be assisted by the rule-extracting feature of FNN.

Small-scale production of Burkholderia cepacia ATCC21808 lipase adapted to high-throughput screening

The screening of variant libraries of recombinant Burkholderia cepacia ATCC21808 lipase generated in Escherichia coli is limited by expression difficulties that are mainly due to the formation of inclusion bodies. To circumvent these difficulties and provide an efficient small-scale screening protocol, the gene encoding the lipase from B. cepacia was expressed in various expression vectors. With the pFLAG-ATS-Lip-Hp construct, expression of up to 6807 U/L of culture was possible in Erlenmeyer flasks. The production protocol was miniaturized in 96 deep-well plates, yielding 1300 U/L of lipase in fusion with the FLAG tag. With this protocol, the activity was determined in less than 10 min for a full plate, with a coefficient of variance of about 25%. For validation, 18 mutants constructed by site-directed mutagenesis on position Valine 266 were screened. Nice variations of activity were detected and found to be in agreement with those obtained in Erlenmeyer flask cultures. The protocol enabled the identification of 5 mutants showing enhanced activity toward para-nitrophenyl butyrate.

Better library design: data-driven protein engineering

Data-driven protein engineering is increasingly used as an alternative to rational design and combinatorial engineering because it uses available knowledge to limit library size, while still allowing for the identification of unpredictable substitutions that lead to large effects. Recent advances in computational modeling and bioinformatics, as well as an increasing databank of experiments on functional variants, have led to new strategies to choose particular amino acid residues to vary in order to increase the chances of obtaining a variant protein with the desired property. Strategies for limiting diversity at each position, design of small sub-libraries, and the performance of scouting experiments, have also been developed or even automated, further reducing the library size.

Recent progress in engineering alpha/beta hydrolase-fold family members

The members of the alpha/beta hydrolase-fold family represent a functionally versatile group of enzymes with many important applications in biocatalysis. Given the technical significance of alpha/beta hydrolases in processes ranging from the kinetic resolution of enantiomeric precursors for pharmaceutical compounds to bulk products such as laundry detergent, optimizing and tailoring enzymes for these applications presents an ongoing challenge to chemists, biochemists, and engineers alike. A review of the recent literature on alpha/beta hydrolase engineering suggests that the early successes of "random processes" such as directed evolution are now being slowly replaced by more hypothesis-driven, focused library approaches. These developments reflect a better understanding of the enzymes' structure-function relationship and improved computational resources, which allow for more sophisticated search and prediction algorithms, as well as, in a very practical sense, the realization that bigger is not always better.