Structure-aware Protein Solubility Prediction From Sequence Through Graph Convolutional Network And Predicted Contact Map

Chen, Jianwen; Zheng, Shuangjia; Zhao, Huiying; Yang, Yuedong

doi:10.1101/2020.06.24.169011

Cited by 5 publications

(4 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Graph neural networks (GNNs) were used to extract substrate features. 40 Substrates were considered as molecular graphs G = (ν,ϵ). Each atom v i ∈ ν was represented by a 46-dimensional vector, which was the concatenation of one-hot encodings representing the atom types, degrees of the atom, chirality, hybridization types, number of radical electrons, number of hydrogen atoms attached, explicit valence, implicit valence, and aromaticity of the corresponding atoms.…”

Section: ■ Discussionmentioning

confidence: 99%

Discovery of Toxin-Degrading Enzymes with Positive Unlabeled Deep Learning

Zhang,

Xing,

Liu

et al. 2024

ACS Catal.

View full text Add to dashboard Cite

Identifying functional enzymes for the catalysis of specific biochemical reactions is a major bottleneck in the de novo design of biosynthesis and biodegradation pathways. Conventional methods based on microbial screening and functional metagenomics require long verification periods and incur high experimental costs; recent data-driven methods apply only to a few common substrates. To enable rapid and high-throughput identification of enzymes for complex and less-studied substrates, we propose a robust enzyme's substrate promiscuity prediction model based on positive unlabeled learning. Using this model, we identified 15 new degrading enzymes specific for the mycotoxins ochratoxin A and zearalenone, of which six could degrade >90% mycotoxin content within 3 h. We anticipate that this model will serve as a useful tool for identifying new functional enzymes and understanding the nature of biocatalysis, thereby advancing the fields of synthetic biology, metabolic engineering, and pollutant biodegradation.

show abstract

Section: ■ Discussionmentioning

confidence: 99%

Discovery of Toxin-Degrading Enzymes with Positive Unlabeled Deep Learning

Zhang,

Xing,

Liu

et al. 2024

ACS Catal.

View full text Add to dashboard Cite

show abstract

“…As molecules interact with the target protein, effective embedding will reduce the limit of the small amount of training samples. The protein information could be embedded through sequence, contact map, or 3D structure . Second, experimental costs limit the discussion about direct comparison of meta learning on recently proposed MO models, which can be covered in future work, and the effectiveness of meta learning on many other directions in low-resource drug discovery is yet to be discovered.…”

Section: Discussionmentioning

confidence: 99%

Meta Learning for Low-Resource Molecular Optimization

Wang

Zheng

Chen

et al. 2021

J. Chem. Inf. Model.

Self Cite

View full text Add to dashboard Cite

The goal of molecular optimization (MO) is to discover molecules that acquire improved pharmaceutical properties over a known starting molecule. Despite many recent successes of new approaches for MO, these methods were typically developed for particular properties with rich annotated training examples. Thus, these approaches are difficult to implement in real scenes where only a small amount of pharmaceutical data is usually available due to the expense and significant effort required for the data collection. Here, we propose a new approach, Meta-MO, for molecular optimization with a handful of training samples based on the wellrecognized first-order meta-learning algorithms. By using a set of meta tasks with rich training samples, Meta-MO trains a meta model through the meta-learning optimization and adapts the learned model to new low-resource MO tasks. Meta-MO was shown to consistently outperform several pretraining and multitask training procedures, providing an average improvement in the success rate of 4.3% on a large-scale bioactivity data set with diverse target variations. We also observed that Meta-MO resulted in the best performing models across fine-tuning sets with only dozens of samples. To the best of our knowledge, this is the first study to apply meta learning to MO tasks. More importantly, such a strategy could be further extended to many low-resource scenarios in realworld drug design.

show abstract

“…The protein representation was done from the perspectives of structure and sequence features, respectively. For the structural feature, we considered using a graph to represent the 2D structure of proteins, which has been proven effective for predicting protein solubility in our previous study [26]. In the protein graph model, residues were regarded as nodes and the contact map predicted from the sequence was used as the adjacency matrix.…”

Section: Representation Of Protein and Moleculementioning

confidence: 99%

X-DPI: A structure-aware multi-modal deep learning model for drug-protein interactions prediction

Wang

Zheng²,

Jiang³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Motivation: Identifying the drug-protein interactions (DPIs) is crucial in drug discovery, and a number of machine learning methods have been developed to predict DPIs. Existing methods usually use unrealistic datasets with hidden bias, which will limit the accuracy of virtual screening methods. Meanwhile, most DPIs prediction methods pay more attention to molecular representation but lack effective research on protein representation and high-level associations between different instances. To this end, we presented here a novel structure-aware multi-modal DPIs prediction model, X-DPI, performing on a curated industry-scale benchmark dataset. Results: We built a high-quality benchmark dataset named GalaxyDB for DPIs prediction. This industry-scale dataset along with an unbiased training procedure resulted in a more robust bench-mark study. For informative protein representation, we constructed a structure-aware graph neural network method from the protein sequence by combining predicted contact maps and graph neural networks. Through further integration of structure-based representation and high-level pre-trained embeddings for molecules and proteins, our model captured more effectively the feature representation of the interactions between them. As a result, X-DPI outperformed state-of-the-art DPIs prediction methods and obtained 5.30% Mean Square Error (MSE) improved in the DAVIS dataset and 8.89% area under the curve (AUC) improved in GalaxyDB dataset. Moreover, our model is an interpretable model with the transformer-based interaction mechanism, which can accurately reveal the binding sites between molecule and protein.

show abstract

Structure-aware Protein Solubility Prediction From Sequence Through Graph Convolutional Network And Predicted Contact Map

Cited by 5 publications

References 34 publications

Discovery of Toxin-Degrading Enzymes with Positive Unlabeled Deep Learning

Discovery of Toxin-Degrading Enzymes with Positive Unlabeled Deep Learning

Meta Learning for Low-Resource Molecular Optimization

X-DPI: A structure-aware multi-modal deep learning model for drug-protein interactions prediction

Contact Info

Product

Resources

About