Dataset Information

Selecting molecules with diverse structures and properties by maximizing submodular functions of descriptors learned with graph neural networks.

ABSTRACT: Selecting diverse molecules from unexplored areas of chemical space is one of the most important tasks for discovering novel molecules and reactions. This paper proposes a new approach for selecting a subset of diverse molecules from a given molecular list by using two existing techniques studied in machine learning and mathematical optimization: graph neural networks (GNNs) for learning vector representation of molecules and a diverse-selection framework called submodular function maximization. Our method, called SubMo-GNN, first trains a GNN with property prediction tasks, and then the trained GNN transforms molecular graphs into molecular vectors, which capture both properties and structures of molecules. Finally, to obtain a subset of diverse molecules, we define a submodular function, which quantifies the diversity of molecular vectors, and find a subset of molecular vectors with a large submodular function value. This can be done efficiently by using the greedy algorithm, and the diversity of selected molecules measured by the submodular function value is mathematically guaranteed to be at least 63% of that of an optimal selection. We also introduce a new evaluation criterion to measure the diversity of selected molecules based on molecular properties. Computational experiments confirm that our SubMo-GNN successfully selects diverse molecules from the QM9 dataset regarding the property-based criterion, while performing comparably to existing methods regarding standard structure-based criteria. We also demonstrate that SubMo-GNN with a GNN trained on the QM9 dataset can select diverse molecules even from other MoleculeNet datasets whose domains are different from the QM9 dataset. The proposed method enables researchers to obtain diverse sets of molecules for discovering new molecules and novel chemical reactions, and the proposed diversity criterion is useful for discussing the diversity of molecular libraries from a new property-based perspective.

SUBMITTER: Nakamura T

PROVIDER: S-EPMC8782878 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BACKGROUND: Avian brood parasites and their hosts are involved in complex offence-defense coevolutionary arms races. The most common pair of reciprocal adaptations in these systems is egg discrimination by hosts and egg mimicry by parasites. As mimicry improves, more advanced host adaptations evolve such as decreased intra- and increased interclutch variation in egg appearance to facilitate detection of parasitic eggs. As interclutch variation increases, parasites able to choose hosts matching best their own egg phenotype should be selected, but this requires that parasites know their own egg phenotype and select host nests correspondingly. METHODOLOGY/PRINCIPAL FINDINGS: We compared egg mimicry of common cuckoo Cuculus canorus eggs in naturally parasitized marsh warbler Acrocephalus palustris nests and their nearest unparasitized conspecific neighbors having similar laying dates and nest-site characteristics. Modeling of avian vision and image analyses revealed no evidence that cuckoos parasitize nests where their eggs better match the host eggs. Cuckoo eggs were as good mimics, in terms of background and spot color, background luminance, spotting pattern and egg size, of host eggs in the nests actually exploited as those in the neighboring unparasitized nests. CONCLUSIONS/SIGNIFICANCE: We reviewed the evidence for brood parasites selecting better-matching host egg phenotypes from several relevant studies and argue that such selection probably cannot exist in host-parasite systems where host interclutch variation is continuous and overall low or moderate. To date there is also no evidence that parasites prefer certain egg phenotypes in systems where it should be most advantageous, i.e., when both hosts and parasites lay polymorphic eggs. Hence, the existence of an ability to select host nests to maximize mimicry by brood parasites appears unlikely, but this possibility should be further explored in cuckoo-host systems where the host has evolved discrete egg phenotypes.

Dataset Information

Selecting molecules with diverse structures and properties by maximizing submodular functions of descriptors learned with graph neural networks.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets