TorchDrug: A Powerful and Flexible Machine Learning Platform for Drug Discovery

Zhu, Zhaocheng; Shi, Chence; Zhang, Zuobai; Liu, Shengchao; Xu, Minghao; Yuan, Xinyu; Zhang, Yangtian; Chen, Junkun; Cai, Huiyu; Lu, Jiarui; Ma, Chao; Liu, Runcheng; Xhonneux, Louis-Pascal; Qin, Meng; Tang, J.

doi:10.48550/arxiv.2202.08320

Cited by 18 publications

(20 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The drug compound is represented as a graph using TorchDrug toolkit [42]. The compound graph is denoted as 𝒢 c = (𝒱 c , ε c ) where each node corresponds to a heavy atom (non-hydrogen atom), and has feature and each edge has feature .…”

Section: Tankbind Modelmentioning

confidence: 99%

TANKBind: Trigonometry-Aware Neural NetworKs for Drug-Protein Binding Structure Prediction

Lu¹,

Zhang³

et al. 2022

Preprint

102

View full text Add to dashboard Cite

Illuminating interactions between proteins and small drug molecules is a long-standing challenge in the field of drug discovery. Despite the importance of understanding these interactions, most previous works are limited by hand-designed scoring functions and insufficient conformation sampling. The recently-proposed graph neural network-based methods provides alternatives to predict protein-ligand complex conformation in a one-shot manner. However, these methods neglect the geometric constraints of the complex structure and weaken the role of local functional regions. As a result, they might produce unreasonable conformations for challenging targets and generalize poorly to novel proteins. In this paper, we propose Trigonometry-Aware Neural networKs for binding structure prediction, TANKBind, that builds trigonometry constraint as a vigorous inductive bias into the model and explicitly attends to all possible binding sites for each protein by segmenting the whole protein into functional blocks. We construct novel contrastive losses with local region negative sampling to jointly optimize the binding interaction and affinity. Extensive experiments show substantial performance gains in comparison to state-of-the-art physics-based and deep learning-based methods on commonly-used benchmark datasets for both binding structure and affinity predictions with variant settings. We release our code at https://github.com/luwei0917/TankBind.

show abstract

Section: Tankbind Modelmentioning

confidence: 99%

TANKBind: Trigonometry-Aware Neural NetworKs for Drug-Protein Binding Structure Prediction

Lu¹,

Zhang³

et al. 2022

Preprint

102

View full text Add to dashboard Cite

show abstract

“…In this section, we introduce the details of our experiments. All these methods are developed based on PyTorch and TorchDrug (Zhu et al, 2022).…”

Section: Methodsmentioning

confidence: 99%

Physics-Inspired Protein Encoder Pre-Training via Siamese Sequence-Structure Diffusion Trajectory Prediction

Zhang¹,

Xu²,

Lozano³

et al. 2023

Preprint

View full text Add to dashboard Cite

Pre-training methods on proteins are recently gaining interest, leveraging either protein sequences or structures, while modeling their joint energy landscape is largely unexplored. In this work, inspired by the success of denoising diffusion models, we propose the DiffPreT approach to pretrain a protein encoder by sequence-structure multimodal diffusion modeling. DiffPreT guides the encoder to recover the native protein sequences and structures from the perturbed ones along the multimodal diffusion trajectory, which acquires the joint distribution of sequences and structures. Considering the essential protein conformational variations, we enhance DiffPreT by a physicsinspired method called Siamese Diffusion Trajectory Prediction (SiamDiff) to capture the correlation between different conformers of a protein. SiamDiff attains this goal by maximizing the mutual information between representations of diffusion trajectories of structurally-correlated conformers. We study the effectiveness of DiffPreT and SiamDiff on both atom-and residue-level structure-based protein understanding tasks. Experimental results show that the performance of DiffPreT is consistently competitive on all tasks, and SiamDiff achieves new state-of-the-art performance, considering the mean ranks on all tasks. The source code will be released upon acceptance.

show abstract

“…Build on top of the PyTorch library, TorchDrug benchmarks a variety of important tasks in drug discovery, including molecular property prediction, pretrained molecular representations, de novo molecular design and optimization, retrosynthsis prediction, and biomedical knowledge graph reasoning (Zhu et al, 2022). Therapeutics Data Commons (TDC) a platform comprised of three components, namely, 66 AI-ready datasets and 22 learning tasks for drug discovery and development, an ecosystem of tools and community resources and leaderboards for therapeutics machine learning (Huang et al, 2022).…”

Section: Deepchem and Torchdrug And Other Chemical Library Examplesmentioning

confidence: 99%

Digital Research Environment(DRE)-enabled Artificial Intelligence (AI) to facilitate early stage drug development

Barrett

Oskoui²,

Russell³

et al. 2023

Front. Pharmacol.

View full text Add to dashboard Cite

Early-stage drug discovery is highly dependent upon drug target evaluation, understanding of disease progression and identification of patient characteristics linked to disease progression overlaid upon chemical libraries of potential drug candidates. Artificial intelligence (AI) has become a credible approach towards dealing with the diversity and volume of data in the modern drug development phase. There are a growing number of services and solutions available to pharmaceutical sponsors though most prefer to constrain their own data to closed solutions given the intellectual property considerations. Newer platforms offer an alternative, outsourced solution leveraging sponsors data with other, external open-source data to anchor predictions (often proprietary algorithms) which are refined given data indexed upon the sponsor’s own chemical libraries. Digital research environments (DREs) provide a mechanism to ingest, curate, integrate and otherwise manage the diverse data types relevant for drug discovery activities and also provide workspace services from which target sharing and collaboration can occur providing yet another alternative with sponsors being in control of the platform, data and predictive algorithms. Regulatory engagement will be essential in the operationalizing of the various solutions and alternatives; current treatment of drug discovery data may not be adequate with respect to both quality and useability in the future. More sophisticated AI/ML algorithms are likely based on current performance metrics and diverse data types (e.g., imaging and genomic data) will certainly be a more consistent part of the myriad of data types that fuel future AI-based algorithms. This favors a dynamic DRE-enabled environment to support drug discovery.

show abstract

TorchDrug: A Powerful and Flexible Machine Learning Platform for Drug Discovery

Cited by 18 publications

References 14 publications

TANKBind: Trigonometry-Aware Neural NetworKs for Drug-Protein Binding Structure Prediction

TANKBind: Trigonometry-Aware Neural NetworKs for Drug-Protein Binding Structure Prediction

Physics-Inspired Protein Encoder Pre-Training via Siamese Sequence-Structure Diffusion Trajectory Prediction

Digital Research Environment(DRE)-enabled Artificial Intelligence (AI) to facilitate early stage drug development

Contact Info

Product

Resources

About