Performance and robustness of small molecule retention time prediction with molecular graph neural networks in industrial drug discovery campaigns.

This study explores how machine-learning can be used to predict chromatographic retention times (RT) for the analysis of small molecules, with the objective of identifying a machine-learning framework with the robustness required to support a chemical synthesis production platform. We used internally generated data from high-throughput parallel synthesis in context of pharmaceutical drug discovery projects. We tested machine-learning models from the following frameworks: XGBoost, ChemProp, and DeepChem, using a dataset of 7552 small molecules. Our findings show that two specific models, AttentiveFP and ChemProp, performed better than XGBoost and a regular neural network in predicting RT accurately. We also assessed how well these models performed over time and found that molecular graph neural networks consistently gave accurate predictions for new chemical series. In addition, when we applied ChemProp on the publicly available METLIN SMRT dataset, it performed impressively with an average error of 38.70 s. These results highlight the efficacy of molecular graph neural networks, especially ChemProp, in diverse RT prediction scenarios, thereby enhancing the efficiency of chromatographic analysis.

Dynamic Modeling of Indole Glucosinolate Hydrolysis and Its Impact on Auxin Signaling.

Plants release chemicals to deter attackers. Arabidopsis thaliana relies on multiple defense compounds, including indol-3-ylmethyl glucosinolate (I3G), which upon hydrolysis initiated by myrosinase enzymes releases a multitude of bioactive compounds, among others, indole-3-acetonitrile and indole-3-acetoisothiocyanate. The highly unstable isothiocyanate rapidly reacts with other molecules. One of the products, indole-3-carbinol, was reported to inhibit auxin signaling through binding to the TIR1 auxin receptor. On the contrary, the nitrile product of I3G hydrolysis can be converted by nitrilase enzymes to form the primary auxin molecule, indole-3-acetic acid, which activates TIR1. This suggests that auxin signaling is subject to both antagonistic and protagonistic effects of I3G hydrolysis upon attack. We hypothesize that I3G hydrolysis and auxin signaling form an incoherent feedforward loop and we build a mathematical model to examine the regulatory network dynamics. We use molecular docking to investigate the possible antagonistic properties of different I3G hydrolysis products by competitive binding to the TIR1 receptor. Our simulations reveal an uncoupling of auxin concentration and signaling, and we determine that enzyme activity and antagonist binding affinity are key parameters for this uncoupling. The molecular docking predicts that several I3G hydrolysis products strongly antagonize auxin signaling. By comparing a tissue disrupting attack - e.g., by chewing insects or necrotrophic pathogens that causes rapid release of I3G hydrolysis products - to sustained cell-autonomous I3G hydrolysis, e.g., upon infection by biotrophic pathogens, we find that each scenario gives rise to distinct auxin signaling dynamics. This suggests that plants have different defense versus growth strategies depending on the nature of the attack.

Unravelling Protein-Protein Interaction Networks Linked to Aliphatic and Indole Glucosinolate Biosynthetic Pathways in Arabidopsis.

Within the cell, biosynthetic pathways are embedded in protein-protein interaction networks. In Arabidopsis, the biosynthetic pathways of aliphatic and indole glucosinolate defense compounds are well-characterized. However, little is known about the spatial orchestration of these enzymes and their interplay with the cellular environment. To address these aspects, we applied two complementary, untargeted approaches-split-ubiquitin yeast 2-hybrid and co-immunoprecipitation screens-to identify proteins interacting with CYP83A1 and CYP83B1, two homologous enzymes specific for aliphatic and indole glucosinolate biosynthesis, respectively. Our analyses reveal distinct functional networks with substantial interconnection among the identified interactors for both pathway-specific markers, and add to our knowledge about how biochemical pathways are connected to cellular processes. Specifically, a group of protein interactors involved in cell death and the hypersensitive response provides a potential link between the glucosinolate defense compounds and defense against biotrophic pathogens, mediated by protein-protein interactions.

CB5C affects the glucosinolate profile in Arabidopsis thaliana.

Cytochrome b5 (CB5) proteins are small heme-binding proteins, that influence cytochrome P450 activity. While only one CB5 isoform is found in mammals, higher plants have several isoforms of these proteins. The roles of the many CB5 isoforms in plants remain unknown. We hypothesized that CB5 proteins support the cytochrome P450 enzymes of plant specialized metabolism and found CB5C from Arabidopsis thaliana to co-express with glucosinolate biosynthetic genes. We characterized the glucosinolate profiles of 2 T-DNA insertion mutants of CB5C, and found that long-chained aliphatic glucosinolates were reduced in one of the mutant lines - a phenotype that was exaggerated upon methyl-jasmonate treatment. These results support the hypothesis, that CB5C influences glucosinolate biosynthesis, however, the mode of action remains unknown. Furthermore, the mutants differed in their biomass response to methyl jasmonate treatment. Thereby, our results highlight the varying effects of T-DNA insertion sites, as the 2 analyzed alleles show different phenotypes.