DeepGraphMolGen, a multi-objective, computational strategy for generating molecules with desirable properties: a graph convolution and reinforcement learning approach

Abstract

We address the problem of generating novel molecules with desired interaction properties as a multi-objective optimization problem. Interaction binding models are learned from binding data using graph convolution networks (GCNs). Since the experimentally obtained property scores are recognised as having potentially gross errors, we adopted a robust loss for the model. Combinations of these terms, including drug likeness and synthetic accessibility, are then optimized using reinforcement learning based on a graph convolution policy approach. Some of the molecules generated, while legitimate chemically, can have excellent drug-likeness scores but appear unusual. We provide an example based on the binding potency of small molecules to dopamine transporters. We extend our method successfully to use a multi-objective reward function, in this case for generating novel molecules that bind with dopamine transporters but not with those for norepinephrine. Our method should be generally applicable to the generation in silico of molecules with desirable properties.

Permutation Equivariant Generative Adversarial Networks for Graphs

“…We use RDKit only for preprocessing and final evaluation metric calculation. Therefore, the comparison with other models is not fair 6 .…”

“…Some use a canonical sequential representation such as the SMILES representation (for instance: [20,17,5,7,2]). Others are generating directly some graph objects such nodes or edges [25,18,9,21,10,8,6].…”

“…Typically, MPNN uses the topological information as masking, setting h to 0, so that the aggregation function is over the neighborhood of node i rather than over all nodes in the graph 5. To do so, we encode the formal charge and the number of hydrogen into the node attribute, so that we have 21 atom types instead of usually 46 In addition, we note that both MolGAN and GG-GAN help their discriminators by adding various handcrafted features. On our side, the use of skeleton directly from the data is a significant advantage.…”

CoPriNet: Deep learning compound price prediction for use in de novo molecule generation and prioritization.

“…Due primarily to their simplicity and speed, SAscore and SCScore have been used extensively across drug development pipelines including for compound screening (e.g., Omolabi et al, 2021;Basu et al, 2020;Lu and Li, 2021;Huang et al, 2019), dataset preparation (e.g., Imrie et al, 2021b;Humbeck et al, 2018) and molecule generation/optimization (e.g., Leguy et al, 2020;Zhou et al, 2019;Khemchandani et al, 2020a;Green et al, 2020). SAScore is one of the most popular metrics for biasing or discarding potentially infeasible compounds in methods for computational generation of de novo molecules (e.g., Yassine et al, 2021;Imrie et al, 2020;Prykhodko et al, 2019;Leguy et al, 2020;Khemchandani et al, 2020b). However, as described above, SAscore and SCscore are simple approximations for SA and as such, present several limitations.…”

Geometric learning of the conformational dynamics of molecules using dynamic graph neural networks

“…Another important task is the generation of molecular graphs with optimized properties [2,13,20,33]. Reinforcement learning, in particular, has been combined with autoencoders [6,12], graph convolutional networks (GCNs) [14,32], or flow models [22,27] to generate static optimized molecular graphs for a variety of applications.…”