Machine Learning to Predict Enzyme-Substrate Interactions in Elucidation of Synthesis Pathways: A Review.

Enzyme-substrate interactions play a fundamental role in elucidating synthesis pathways and synthetic biology, as they allow for the understanding of important aspects of a reaction. Establishing the interaction experimentally is a slow and costly process, which is why this problem has been addressed using computational methods such as molecular dynamics, molecular docking, and Monte Carlo simulations. Nevertheless, this type of method tends to be computationally slow when dealing with a large search space. Therefore, in recent years, methods based on artificial intelligence, such as support vector machines, neural networks, or decision trees, have been implemented, significantly reducing the computing time and covering vast search spaces. These methods significantly reduce the computation time and cover broad search spaces, rapidly reducing the number of interacting candidates, as they allow repetitive processes to be automated and patterns to be extracted, are adaptable, and have the capacity to handle large amounts of data. This article analyzes these artificial intelligence-based approaches, presenting their common structure, advantages, disadvantages, limitations, challenges, and future perspectives.

Genome-Scale Metabolic Reconstruction, Non-Targeted LC-QTOF-MS Based Metabolomics Data, and Evaluation of Anticancer Activity of Cannabis sativa Leaf Extracts.

Over the past decades, Colombia has suffered complex social problems related to illicit crops, including forced displacement, violence, and environmental damage, among other consequences for vulnerable populations. Considerable effort has been made in the regulation of illicit crops, predominantly Cannabis sativa, leading to advances such as the legalization of medical cannabis and its derivatives, the improvement of crops, and leaving an open window to the development of scientific knowledge to explore alternative uses. It is estimated that C. sativa can produce approximately 750 specialized secondary metabolites. Some of the most relevant due to their anticancer properties, besides cannabinoids, are monoterpenes, sesquiterpenoids, triterpenoids, essential oils, flavonoids, and phenolic compounds. However, despite the increase in scientific research on the subject, it is necessary to study the primary and secondary metabolism of the plant and to identify key pathways that explore its great metabolic potential. For this purpose, a genome-scale metabolic reconstruction of C. sativa is described and contextualized using LC-QTOF-MS metabolic data obtained from the leaf extract from plants grown in the region of Pesca-Boyaca, Colombia under greenhouse conditions at the Clever Leaves facility. A compartmentalized model with 2101 reactions and 1314 metabolites highlights pathways associated with fatty acid biosynthesis, steroids, and amino acids, along with the metabolism of purine, pyrimidine, glucose, starch, and sucrose. Key metabolites were identified through metabolomic data, such as neurine, cannabisativine, cannflavin A, palmitoleic acid, cannabinoids, geranylhydroquinone, and steroids. They were analyzed and integrated into the reconstruction, and their potential applications are discussed. Cytotoxicity assays revealed high anticancer activity against gastric adenocarcinoma (AGS), melanoma cells (A375), and lung carcinoma cells (A549), combined with negligible impact against healthy human skin cells.

Economic evaluation of the development of a phage therapy product for the control of Salmonella in poultry.

Poultry products are one of the major transmission media of Salmonella enteritidis to humans. A promising alternative to reduce the load of Salmonella in poultry are bacteriophages. Elsewhere, a mixture of six bacteriophages has been used successfully, but large-scale production would be necessary to supply potential poultry market and costs analyses have not been calculated yet. For this, a powerful tool to predict production costs is bioprocess modeling coupled with economic analyses. This work aims to model the scaled-up production of a six bacteriophages mixture based on a laboratory/pilot-scale production using Biosolve Process. For the model construction, a combination of experimental and reported data was applied, in which different production alternatives and the range of 1-100% of the Colombian poultry market (at broiler's farm and slaughterhouse) were analyzed. Results indicate that the best cost-effective process configuration/scale is to use one bioreactor (156 L) for the six bacteriophages, then a 0.45 µm filtration for removal of biomass, and a 0.22 µm filtration for sterility; this to supply the 35% of the market size for broiler farms (equivalent to 210 million chickens). This configuration gives a production cost per chicken of US$ 0.02. Additionally, a sensitivity analysis and a theoretical contrast for understanding the impact that titer and recovery have on production scale determined that titer affects the most the cost and requires optimization. The present works serves as a first, and required, approach for the development of phage therapy products that are alternatives to present-day pathogens control strategies.

Lipid Metabolic Versatility in Malassezia spp. Yeasts Studied through Metabolic Modeling.

Malassezia species are lipophilic and lipid-dependent yeasts belonging to the human and animal microbiota. Typically, they are isolated from regions rich in sebaceous glands. They have been associated with dermatological diseases such as seborrheic dermatitis, pityriasis versicolor, atopic dermatitis, and folliculitis. The genomes of Malassezia globosa, Malassezia sympodialis, and Malassezia pachydermatis lack the genes related to fatty acid synthesis. Here, the lipid-synthesis pathways of these species, as well as of Malassezia furfur, and of an atypical M. furfur variant were reconstructed using genome data and Constraints Based Reconstruction and Analysis. To this end, the genomes of M. furfur CBS 1878 and the atypical M. furfur 4DS were sequenced and annotated. The resulting Enzyme Commission numbers and predicted reactions were similar to the other Malassezia strains despite the differences in their genome size. Proteomic profiling was utilized to validate flux distributions. Flux differences were observed in the production of steroids in M. furfur and in the metabolism of butanoate in M. pachydermatis. The predictions obtained via these metabolic reconstructions also suggested defects in the assimilation of palmitic acid in M. globosa, M. sympodialis, M. pachydermatis, and the atypical variant of M. furfur, but not in M. furfur. These predictions were validated via physiological characterization, showing the predictive power of metabolic network reconstructions to provide new clues about the metabolic versatility of Malassezia.

A network model for biofilm development in Escherichia coli K-12.

BACKGROUND: In nature, bacteria often exist as biofilms. Biofilms are communities of microorganisms attached to a surface. It is clear that biofilm-grown cells harbor properties remarkably distinct from planktonic cells. Biofilms frequently complicate treatments of infections by protecting bacteria from the immune system, decreasing antibiotic efficacy and dispersing planktonic cells to distant body sites. In this work, we employed enhanced Boolean algebra to model biofilm formation. RESULTS: The network obtained describes biofilm formation successfully, assuming - in accordance with the literature - that when the negative regulators (RscCD and EnvZ/OmpR) are off, the positive regulator (FlhDC) is on. The network was modeled under three different conditions through time with satisfactory outcomes. Each cluster was constructed using the K-means/medians Clustering Support algorithm on the basis of published Affymetrix microarray gene expression data from biofilm-forming bacteria and the planktonic state over four time points for Escherichia coli K-12. CONCLUSIONS: The different phenotypes obtained demonstrate that the network model of biofilm formation can simulate the formation or repression of biofilm efficiently in E. coli K-12.

Escherichia coli autoinducer-2 uptake network does not display hysteretic behavior but AI-2 synthesis rate controls transient bifurcation.

Analysis of different architectures of quorum sensing networks has been the center of attention in recent times. The approach employs mathematical models to uncover the factors behind the dynamics. Quorum sensing networks mostly display autoregulation such as Pseudomonas aeruginosa and Vibrio cholerae. However, Escherichia coli autoinducer-2 (AI-2) synthesis does not display autoinduction (i.e. autoregulation). This and other features have raised questions about the actual function of AI-2 inside the cell. In this paper we propose a model for lsr operon regulation which explains or at least is consistent with AI-2 uptake in E. coli. The model was employed to determine the main factors that control the concentration of the signal and the uptake activation. We investigated deterministic and stochastic variants of the network model and we found no states that could lead to the typical bistability in quorum sensing systems. However, stochastic simulations predict a transient bifurcation (positively regulated by AI-2 synthesis) that could provide some advantage in adapting to new environments. LsrR inactivation was found to play a crucial role in the uptake activation compared to AI-2 synthesis, lsr transcription and AI-2 excretion. Our hypothesis is that positive regulation of the level of expression is the main factor in understanding the function of the lsr operon. This is in contrast to the conventionally held belief that the main factor is the onset of activation.

Motility influences biofilm architecture in Escherichia coli.

Eight Escherichia coli strains were studied in minimal medium with a continuous flow system using confocal microscopy. K12 wild-type strains ATCC 25404 and MG1655 formed the best biofilms ( approximately 43 microm thick, 21 to 34% surface coverage). JM109, DH5alpha, and MG1655 motA formed intermediate biofilms ( approximately 13 microm thick, 41 to 58% surface coverage). BW25113, MG1655 qseB, and MG1655 fliA had poor biofilms (surface coverage less than 5%). The best biofilm-formers, ATCC 25404 and MG1655, displayed the highest motility, whereas the worst biofilm former, BW25113, was motility-impaired. The differences in motility were due to differences in expression of the motility loci qseB, flhD, fliA, fliC, and motA (e.g., qseB expression in MG1655 was 139-fold higher than BW25113 and 209-fold higher than JM109). Motility affected the biofilm architecture as those strains which had poor motility (E. coli JM109, E. coli MG1655 motA, and DH5alpha) formed flatter microcolonies compared with MG1655 and ATCC 25404, which had more dramatic vertical structures as a result of their enhanced motility. The presence of flagella was also found to be important as qseB and fliA mutants (which lack flagella) had less biofilm than the isogenic paralyzed motA strain (threefold less thickness and 15-fold less surface coverage).

Autoinducer 2 controls biofilm formation in Escherichia coli through a novel motility quorum-sensing regulator (MqsR, B3022).

The cross-species bacterial communication signal autoinducer 2 (AI-2), produced by the purified enzymes Pfs (nucleosidase) and LuxS (terminal synthase) from S-adenosylhomocysteine, directly increased Escherichia coli biofilm mass 30-fold. Continuous-flow cells coupled with confocal microscopy corroborated these results by showing the addition of AI-2 significantly increased both biofilm mass and thickness and reduced the interstitial space between microcolonies. As expected, the addition of AI-2 to cells which lack the ability to transport AI-2 (lsr null mutant) failed to stimulate biofilm formation. Since the addition of AI-2 increased cell motility through enhanced transcription of five motility genes, we propose that AI-2 stimulates biofilm formation and alters its architecture by stimulating flagellar motion and motility. It was also found that the uncharacterized protein B3022 regulates this AI-2-mediated motility and biofilm phenotype through the two-component motility regulatory system QseBC. Deletion of b3022 abolished motility, which was restored by expressing b3022 in trans. Deletion of b3022 also decreased biofilm formation significantly, relative to the wild-type strain in three media (46 to 74%) in 96-well plates, as well as decreased biomass (8-fold) and substratum coverage (19-fold) in continuous-flow cells with minimal medium (growth rate not altered and biofilm restored by expressing b3022 in trans). Deleting b3022 changed the wild-type biofilm architecture from a thick (54-mum) complex structure to one that contained only a few microcolonies. B3022 positively regulates expression of qseBC, flhD, fliA, and motA, since deleting b3022 decreased their transcription by 61-, 25-, 2.4-, and 18-fold, respectively. Transcriptome analysis also revealed that B3022 induces crl (26-fold) and flhCD (8- to 27-fold). Adding AI-2 (6.4 muM) increased biofilm formation of wild-type K-12 MG1655 but not that of the isogenic b3022, qseBC, fliA, and motA mutants. Adding AI-2 also increased motA transcription for the wild-type strain but did not stimulate motA transcription for the b3022 and qseB mutants. Together, these results indicate AI-2 induces biofilm formation in E. coli through B3022, which then regulates QseBC and motility; hence, b3022 has been renamed the motility quorum-sensing regulator gene (the mqsR gene).

Differential gene expression for investigation of Escherichia coli biofilm inhibition by plant extract ursolic acid.

After 13,000 samples of compounds purified from plants were screened, a new biofilm inhibitor, ursolic acid, has been discovered and identified. Using both 96-well microtiter plates and a continuous flow chamber with COMSTAT analysis, 10 microg of ursolic acid/ml inhibited Escherichia coli biofilm formation 6- to 20-fold when added upon inoculation and when added to a 24-h biofilm; however, ursolic acid was not toxic to E. coli, Pseudomonas aeruginosa, Vibrio harveyi, and hepatocytes. Similarly, 10 microg of ursolic acid/ml inhibited biofilm formation by >87% for P. aeruginosa in both complex and minimal medium and by 57% for V. harveyi in minimal medium. To investigate the mechanism of this nontoxic inhibition on a global genetic basis, DNA microarrays were used to study the gene expression profiles of E. coli K-12 grown with or without ursolic acid. Ursolic acid at 10 and 30 microg/ml induced significantly (P < 0.05) 32 and 61 genes, respectively, and 19 genes were consistently induced. The consistently induced genes have functions for chemotaxis and mobility (cheA, tap, tar, and motAB), heat shock response (hslSTV and mopAB), and unknown functions (such as b1566 and yrfHI). There were 31 and 17 genes repressed by 10 and 30 microg of ursolic acid/ml, respectively, and 12 genes were consistently repressed that have functions in cysteine synthesis (cysK) and sulfur metabolism (cysD), as well as unknown functions (such as hdeAB and yhaDFG). Ursolic acid inhibited biofilms without interfering with quorum sensing, as shown with the V. harveyi AI-1 and AI-2 reporter systems. As predicted by the differential gene expression, deleting motAB counteracts ursolic acid inhibition (the paralyzed cells no longer become too motile). Based on the differential gene expression, it was also discovered that sulfur metabolism (through cysB) affects biofilm formation (in the absence of ursolic acid).