Simple GNN Regularisation for 3D Molecular Property Prediction &amp; Beyond

Godwin, Jonathan; Schaarschmidt, Michael; Gaunt, Alexander L.; Sánchez‐González, Álvaro; Rubanova, Yulia; Veličković, Petar; Kirkpatrick, Charles James; Battaglia, Peter W.

doi:10.48550/arxiv.2106.07971

Cited by 12 publications

(16 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For our GNN architecture, we adopt the Encode-Process-Decode logic that is now popular for learning on graph-based physics problems Godwin et al, 2022), albeit with some notable modifications. Figure 3 shows an overview of the Flow Reconstruction GNN.…”

Section: Architecturementioning

confidence: 99%

Graph Neural Networks for Aerodynamic Flow Reconstruction from Sparse Sensing

Duthé¹,

Abdallah²,

Barber³

et al. 2023

Preprint

View full text Add to dashboard Cite

Sensing the fluid flow around an arbitrary geometry entails extrapolating from the physical quantities perceived at its surface in order to reconstruct the features of the surrounding fluid. This is a challenging inverse problem, yet one that if solved could have a significant impact on many engineering applications. The exploitation of such an inverse logic has gained interest in recent years with the advent of widely available cheap but capable MEMS-based sensors. When combined with novel data-driven methods, these sensors may allow for flow reconstruction around immersed structures, benefiting applications such as unmanned airborne/underwater vehicle path planning or control and structural health monitoring of wind turbine blades. In this work, we train deep reversible Graph Neural Networks (GNNs) to perform flow sensing (flow reconstruction) around two-dimensional aerodynamic shapes: airfoils. Motivated by recent work, which has shown that GNNs can be powerful alternatives to mesh-based forward physics simulators, we implement a Message-Passing Neural Network to simultaneously reconstruct both the pressure and velocity fields surrounding simulated airfoils based on their surface pressure distributions, whilst additionally gathering useful farfield properties in the form of context vectors. We generate a unique dataset of Computational Fluid Dynamics simulations by simulating random, yet meaningful combinations of input boundary conditions and airfoil shapes. We show that despite the challenges associated with reconstructing the flow around arbitrary airfoil geometries in high Reynolds turbulent inflow conditions, our framework is able to generalize well to unseen cases.

show abstract

Section: Architecturementioning

confidence: 99%

Graph Neural Networks for Aerodynamic Flow Reconstruction from Sparse Sensing

Duthé¹,

Abdallah²,

Barber³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…As demonstrated by Wu, Zhang, Jin, Jiang, and Li, [79] the denoising diffusion architecture [80][81][82] had a strong connectivity with the enhanced sampling method in MD, [83][84][85][86] where energy was injected into the microscopic system to smooth biomolecular potential energy surface and decrease energy barriers. Besides, it had been shown in Godwin et al [33] that the simple noise regularization could be an effective way to address oversmoothing. [87] A noise correction target could be added to prevent oversmoothing by enforcing diversity in the last few layers of the GNN, which was implemented with an auxiliary denoising autoencoder loss.…”

Section: Time-series Prompting For Motion Predictionmentioning

confidence: 99%

Pre‐Training of Equivariant Graph Matching Networks with Conformation Flexibility for Drug Binding

Jin

Jiang³

et al. 2022

Advanced Science

View full text Add to dashboard Cite

The latest biological findings observe that the motionless “lock‐and‐key” theory is not generally applicable and that changes in atomic sites and binding pose can provide important information for understanding drug binding. However, the computational expenditure limits the growth of protein trajectory‐related studies, thus hindering the possibility of supervised learning. A spatial‐temporal pre‐training method based on the modified equivariant graph matching networks, dubbed ProtMD which has two specially designed self‐supervised learning tasks: atom‐level prompt‐based denoising generative task and conformation‐level snapshot ordering task to seize the flexibility information inside molecular dynamics (MD) trajectories with very fine temporal resolutions is presented. The ProtMD can grant the encoder network the capacity to capture the time‐dependent geometric mobility of conformations along MD trajectories. Two downstream tasks are chosen to verify the effectiveness of ProtMD through linear detection and task‐specific fine‐tuning. A huge improvement from current state‐of‐the‐art methods, with a decrease of 4.3% in root mean square error for the binding affinity problem and an average increase of 13.8% in the area under receiver operating characteristic curve and the area under the precision‐recall curve for the ligand efficacy problem is observed. The results demonstrate a strong correlation between the magnitude of conformation's motion in the 3D space and the strength with which the ligand binds with its receptor.

show abstract

“…To mitigate the over-smoothing problem, we used multiple aggregation functions through the message-passing scheme, [45] along with residual connections. [46,47] The encoder is constructed by stacking several propagation layers followed by a ReLU activation function. Node embeddings are then concatenated with a residual connection from the input graph.…”

Section: Lone Pair Predictionmentioning

confidence: 99%

“…Multiple solutions were proposed: residual connections augmentations. [46,47] In this work, we concatenate outputs of intermediate layers tackling both the over smoothing and vanishing gradients problems.…”

Section: Learning Natural Bond Interactionsmentioning

confidence: 99%

Stereoelectronics-Aware Molecular Representation Learning

Boiko

Reschützegger

Sánchez-Lengeling

et al. 2022

Preprint

View full text Add to dashboard Cite

The representation of molecular structures is crucial for molecular machine learning strategies. Although graph representations are highly versatile and show their broad applicability, they lack information about the quantum-chemical properties of molecular structures. This work proposes a new way to infuse such information into molecular graphs, using a supervised learning method. As a result, the model is able to predict essential higher-order interactions between electron-rich and electron-deficient localized orbitals. The learned interactions are then used as a representation for the prediction of downstream tasks, improving over QM9 baselines.

show abstract

Simple GNN Regularisation for 3D Molecular Property Prediction & Beyond

Cited by 12 publications

References 0 publications

Graph Neural Networks for Aerodynamic Flow Reconstruction from Sparse Sensing

Graph Neural Networks for Aerodynamic Flow Reconstruction from Sparse Sensing

Pre‐Training of Equivariant Graph Matching Networks with Conformation Flexibility for Drug Binding

Stereoelectronics-Aware Molecular Representation Learning

Contact Info

Product

Resources

About