Self-supervised learning with chemistry-aware fragmentation for effective molecular property prediction.

Molecular property prediction (MPP) is a crucial and fundamental task for AI-aided drug discovery (AIDD). Recent studies have shown great promise of applying self-supervised learning (SSL) to producing molecular representations to cope with the widely-concerned data scarcity problem in AIDD. As some specific substructures of molecules play important roles in determining molecular properties, molecular representations learned by deep learning models are expected to attach more importance to such substructures implicitly or explicitly to achieve better predictive performance. However, few SSL pre-trained models for MPP in the literature have ever focused on such substructures. To challenge this situation, this paper presents a Chemistry-Aware Fragmentation for Effective MPP (CAFE-MPP in short) under the self-supervised contrastive learning framework. First, a novel fragment-based molecular graph (FMG) is designed to represent the topological relationship between chemistry-aware substructures that constitute a molecule. Then, with well-designed hard negative pairs, a is pre-trained on fragment-level by contrastive learning to extract representations for the nodes in FMGs. Finally, a Graphormer model is leveraged to produce molecular representations for MPP based on the embeddings of fragments. Experiments on 11 benchmark datasets show that the proposed CAFE-MPP method achieves state-of-the-art performance on 7 of the 11 datasets and the second-best performance on 3 datasets, compared with six remarkable self-supervised methods. Further investigations also demonstrate that CAFE-MPP can learn to embed molecules into representations implicitly containing the information of fragments highly correlated to molecular properties, and can alleviate the over-smoothing problem of graph neural networks.

Molecular property prediction by semantic-invariant contrastive learning.

MOTIVATION: Contrastive learning has been widely used as pretext tasks for self-supervised pre-trained molecular representation learning models in AI-aided drug design and discovery. However, existing methods that generate molecular views by noise-adding operations for contrastive learning may face the semantic inconsistency problem, which leads to false positive pairs and consequently poor prediction performance. RESULTS: To address this problem, in this article, we first propose a semantic-invariant view generation method by properly breaking molecular graphs into fragment pairs. Then, we develop a Fragment-based Semantic-Invariant Contrastive Learning (FraSICL) model based on this view generation method for molecular property prediction. The FraSICL model consists of two branches to generate representations of views for contrastive learning, meanwhile a multi-view fusion and an auxiliary similarity loss are introduced to make better use of the information contained in different fragment-pair views. Extensive experiments on various benchmark datasets show that with the least number of pre-training samples, FraSICL can achieve state-of-the-art performance, compared with major existing counterpart models. AVAILABILITY AND IMPLEMENTATION: The code is publicly available at https://github.com/ZiqiaoZhang/FraSICL.

Can Pretrained Models Really Learn Better Molecular Representations for AI-Aided Drug Discovery?

Self-supervised pretrained models are gaining increasingly more popularity in AI-aided drug discovery, leading to more and more pretrained models with the promise that they can extract better feature representations for molecules. Yet, the quality of learned representations has not been fully explored. In this work, inspired by the two phenomena of Activity Cliffs (ACs) and Scaffold Hopping (SH) in traditional Quantitative Structure-Activity Relationship analysis, we propose a method named Representation-Property Relationship Analysis (RePRA) to evaluate the quality of the representations extracted by the pretrained model and visualize the relationship between the representations and properties. The concepts of ACs and SH are generalized from the structure-activity context to the representation-property context, and the underlying principles of RePRA are analyzed theoretically. Two scores are designed to measure the generalized ACs and SH detected by RePRA, and therefore, the quality of representations can be evaluated. In experiments, representations of molecules from 10 target tasks generated by 7 pretrained models are analyzed. The results indicate that the state-of-the-art pretrained models can overcome some shortcomings of canonical Extended-Connectivity FingerPrints, while the correlation between the basis of the representation space and specific molecular substructures are not explicit. Thus, some representations could be even worse than the canonical fingerprints. Our method enables researchers to evaluate the quality of molecular representations generated by their proposed self-supervised pretrained models. And our findings can guide the community to develop better pretraining techniques to regularize the occurrence of ACs and SH.

Chemical Composition, Antibacterial Activity and Mechanism of Action of Fermentation Products from Aspergillus Niger xj.

Six compounds were isolated and purified from the crude acetone extract of Aspergillus niger xj. Characterization of all compounds was done by NMR and MS. On the basis of chemical and spectral analysis structure, six compounds were elucidated as metazachlor (1), nonacosane (2), palmitic acid (3), 5,5'-oxybis(5-methylene-2-furaldehyde) (4), dimethyl 5-nitroisophthalate (5) and cholesta-3,5-dien-7-one (6), respectively, and compounds 1, 4, 5 and 6 were isolated for the first time from A. niger. To evaluate the antibacterial activity of compounds 1-6 against three plant pathogenic bacteria (Agrobacterium tumefaciens T-37, Erwinia carotovora EC-1 and Ralstonia solanacearum RS-2), and the minimum inhibitory concentrations (MICs) were determined by broth microdilution method in 96-well microtiter plates. Results of the evaluation of the antibacterial activity showed that T-37 strain was more susceptible to metazachlor with the lowest MIC of 31.25 µg/mL. The antibacterial activity of metazachlor has rarely been reported, thus the antibacterial mechanism of metazachlor against T-37 strain were investigated. The permeability of cell membrane demonstrated that cells membranes were broken by metazachlor, which caused leakage of ions in cells. SDS-PAGE of T-37 proteins indicated that metazachlor could damage bacterial cells through the destruction of cellular proteins. Scanning electron microscopy results showed obvious morphological and ultrastructural changes in the T-37 cells, further confirming the cell membrane damages caused by metazachlor. Overall, our findings demonstrated that the ability of metazachlor to suppress the growth of T-37 pathogenic bacteria makes it potential biocontrol agents.

The antibacterial mechanism of phenylacetic acid isolated from Bacillus megaterium L2 against Agrobacterium tumefaciens.

Background: Agrobacterium tumefaciens T-37 can infect grapes and other fruit trees and cause root cancer. Given the pollution and damage of chemical agents to the environment, the use of biological control has become an important area of focus. Bacillus megaterium L2 is a beneficial biocontrol strain isolated and identified in the laboratory, which has a good antibacterial effect on a variety of plant pathogens. The antibacterial metabolites of L2 were separated and purified to obtain a bioactive compound phenylacetic acid (PAA). Methods: The potential antibacterial mechanism of PAA against A. tumefaciens T-37 strain was determined by relative conductivity, leakage of nucleic acids, proteins, and soluble total sugars, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and reactive oxygen species (ROS). Results: PAA showed good antibacterial activity against strain A. tumefaciens T-37 with IC50 of 0.8038 mg/mL. Our data suggested that after treatment with PAA, the relative conductivity, nucleic acid, protein, and total soluble sugar of T-37 were increased significantly compared with the chloramphenicol treatment group and the negative treatment group. The total protein synthesis of T-37 cells was inhibited, the consumption of phosphorus decreased with the increase of incubation time, and the content of ROS was significantly higher than that in the negative treatment group. Meanwhile, the activity of two key enzymes (MDH and SDH) involved in the tricarboxylic acid cycle (TCA cycle) decreased. In addition, T-37 cells were found to be damaged by scanning electron microscopy observation. Our results showed that PAA can destroy cell membrane integrity, damage cell structures, affect cell metabolism, and inhibit protein synthesis to exert an antibacterial effect. Conclusions: We concluded that the mechanism of action of the PAA against strain T-37 might be described as PAA exerting antibacterial activity by affecting cell metabolism, inhibiting protein synthesis, and destroying cell membrane integrity and cell ultrastructure. Therefore, PAA has a promising application prospect in the prevention and treatment of root cancer disease caused by A. tumefaciens.

Isolation of Bioactive Compounds, Antibacterial Activity, and Action Mechanism of Spore Powder From Aspergillus niger xj.

Aspergillus fungi can produce a wide range of secondary metabolites, and they have represented a potential resource of novel bioactive compounds. Bacterial plant diseases have a serious impact on the sustainable development of agriculture worldwide, so it is necessary to use natural antibacterial compounds in microorganisms to control plant pathogens. This study was conducted to investigate the bioactive compounds of Aspergillus niger xj, three plant pathogens (Agrobacterium tumefaciens T-37, Erwinia carotovora EC-1, and Ralstonia solanacearum RS-2) were used as indicator bacteria, according to the biological activity tracking, five compounds were isolated from A. niger xj spore powder, and characterization of compounds was done by NMR (1H-NMR and 13C-NMR) and EI-MS and was identified as ergosterol (1), ß-sitosterol (2), 5-pentadecylresorcinol (3), 5-hydroxymethyl-2-furancarboxylic acid (4), and succinimide (5). Compounds 3 and 5 were isolated from A. niger xj for the first time. The minimum inhibitory concentration (MIC) of five compounds against three plant pathogens was evaluated, the results showed that compound 4 exhibited the strongest antibacterial activity against tested bacteria, and RS-2 was the most sensitive to compound 4, showing the lowest MIC of 15.56 µg/ml. We concluded that the mechanism of action of the compound 4 against RS-2 might be described as compound 4 acting on bacterial protein synthesis and intracellular metabolism according to the results of the scanning electron microscopy observation, permeability of cell membrane and SDS-PAGE. These results indicated that compound 4 has good potential to be as a biocontrol agent. In conclusion, the results from this study demonstrated that the compounds with antibacterial activity are of great significance of the prevention and control of plant phytopathogenic bacteria, and they may be applicable to exploring alternative approaches to integrated control of phytopathogens.

Mutation breeding of Aspergillus niger by atmospheric room temperature plasma to enhance phosphorus solubilization ability.

Background: Phosphorus (P) is abundant in soils, including organic and inorganic forms. Nevertheless, most of P compounds cannot be absorbed and used by plants. Aspergillus niger v. Tiegh is a strain that can efficiently degrade P compounds in soils. Methods: In this study, A. niger xj strain was mutated using Atmospheric Room Temperature Plasma (ARTP) technology and the strains were screened by Mo-Sb Colorimetry with strong P-solubilizing abilities. Results: Compared with the A. niger xj strain, setting the treatment time of mutagenesis to 120 s, four positive mutant strains marked as xj 90-32, xj120-12, xj120-31, and xj180-22 had higher P-solubilizing rates by 50.3%, 57.5%, 55.9%, and 61.4%, respectively. Among them, the xj120-12 is a highly efficient P solubilizing and growth-promoting strain with good application prospects. The growth characteristics such as plant height, root length, and dry and fresh biomass of peanut (Arachis hypogaea L.) increased by 33.5%, 43.8%, 43.4%, and 33.6%, respectively. Besides available P, the chlorophyll and soluble protein contents also vary degrees of increase in the P-solubilizing mutant strains. Conclusions: The results showed that the ARTP mutagenesis technology can improve the P solubilization abilities of the A. niger mutant strains and make the biomass of peanut plants was enhanced of mutant strains.

Isolation and Identification of Antibacterial Bioactive Compounds From Bacillus megaterium L2.

Bacterial metabolites exhibit a variety of biologically active compounds including antibacterial and antifungal activities. It is well known that Bacillus is considered to be a promising source of bioactive secondary metabolites. Most plant pathogens have an incredible ability to mutate and acquire resistance, causing major economic losses in the agricultural field. Therefore, it is necessary to use the natural antibacterial compounds in microbes to control plant pathogens. This study was conducted to investigate the bio-active compounds of Bacillus megaterium L2. According to the activity guidance of Agrobacterium tumefaciens T-37, Erwinia carotovora EC-1 and Ralstonia solanacearum RS-2, five monomeric compounds, including erucamide (1), behenic acid (2), palmitic acid (3), phenylacetic acid (4), and ß-sitosterol (5), were fractionated and purified from the crude ethyl acetate extract of B. megaterium. To our knowledge, all compounds were isolated from the bacterium for the first time. To understand the antimicrobial activity of these compounds, and their minimum inhibitory concentrations (MICs) (range: 0.98∼500 µg/mL) were determined by the broth microdilution method. For the three tested pathogens, palmitic acid exhibited almost no antibacterial activity (>500 µg/mL), while erucamide had moderate antibacterial activity (MIC = 500 µg/mL). Behenic acid showed MICs of 250 µg/mL against T-37 and RS-2 strains with an antibacterial activity. ß-sitosterol showed significant antimicrobial activity against RS-2. ß-sitosterol showed remarkable antimicrobial activity against RS-2 with an MIC of 15.6 µg/mL. In addition, with the antimicrobial activity, against T-37 (62.5 µg/mL) and against EC-1 (125 µg/mL) and RS-2 (15.6 µg/mL) strains notably, phenylacetic acid may be interesting for the prevention and control of phytopathogenic bacteria. Our findings suggest that isolated compounds such as behenic acid, ß-sitosterol, and phenylacetic acid may be promising candidates for natural antimicrobial agents.