Search

Recently, pre-trained foundation models have shown significant advancements in multiple fields. However, the lack of datasets with labeled features and codebases has hindered the development of a supervised foundation model for molecular tasks. Here, we have carefully curated seven datasets specifically tailored for node- and graph-level prediction tasks to facilitate supervised learning on molecules. Moreover, to support the development of multi-task learning on our proposed datasets, we created the Graphium graph machine learning library. Our dataset collection encompasses two distinct categories. Firstly, the TOYMIX category modifies three small existing datasets with additional data for multi-task learning. Secondly, the LARGEMIX category includes four large-scale datasets with 344M graph-level data points and 409M node-level data points from ∼5M unique molecules. Finally, the ultra-large dataset contains 2,210M graph-level data points and 2,031M node-level data pointscoming from 86M molecules. Hence our datasets represent an order of magnitude increase in data volume compared to other 2D-GNN datasets. In addition, recognizing that molecule-related tasks often span multiple levels, we have designed our library toexplicitly support multi-tasking, offering a diverse range of multi-level representations, i.e., representations at the graph, node, edge, and node-pair level. We equipped the library with an extensive collection of models and features to cover different levels ofmolecule analysis. By combining our curated datasets with this versatile library, we aim to accelerate the development of molecule foundation models. Datasets and code are available at https://github.com/datamol-io/graphium.

Pretraining foundation models that adapt to a wide range of molecule tasks have been long pursued by the community of drug discovery. While self-supervised learning methods are developed to leverage the sheer number of unlabeled molecules for pretraining, the landscape of supervised learning is much underexplored due to the absence of proper datasets and codebases. To facilitate the study of supervised learning on molecules, we curate 7 datasets with node- and graph-level supervision, and develop a library for studying multi-task learning models. The datasets are separated into 2 categories. First, the Toy-mix category contains 3 small datasets that are well known and well studied in the literature, but with the additional constraint that they must be used in a multi-task setting. Second, the Large-mix category contains 4 large datasets that, together, contain tens of billions of graph-level data points and tens of billions node-level data points associated to 100M unique molecules, representing orders of magnitude more data than other 2D-GNN datasets. Since molecule tasks are distributed across multiple levels, we design our library to explicitly consider multi-tasking alongside multi-level representations, backed by a large collection of models and features for different levels. With such a library design to accompany the datasets, we hope to accelerate the development of foundational models for molecules.&nbsp

Antibiotic therapy has drastically reduced the mortality and sequelae of bacterial infections. From naturally occurring to chemically synthesized, different classes of antibiotics have been successfully used without detailed knowledge of how they affect bacterial dynamics in vivo. However, a proportion of patients receiving antimicrobial therapy develop recrudescent infections post-treatment. Relapsing infections are attributable to incomplete clearance of bacterial populations following antibiotic administration; the metabolic profile of this antibiotic-recalcitrant bacterial subpopulation, the spatio-temporal context of its emergence and the variance of antibiotic-bacterial interactions in vivo remain unclear. Here, we develop and apply a mechanistic mathematical model to data from a study comparing the effects of ciprofloxacin and ampicillin on the within-host dynamics of Salmonella enterica serovar Typhimurium in murine infections. Using the inferential capacity of our model, we show that the antibiotic-recalcitrant bacteria following ampicillin, but not ciprofloxacin, treatment belong to a non-replicating phenotype. Aligning with previous studies, we independently estimate that the lymphoid tissues and spleen are important reservoirs of non-replicating bacteria. Finally, we predict that post-treatment, the progenitors of the non-growing and growing bacterial populations replicate and die at different rates. Ultimately, the liver, spleen and mesenteric lymph nodes are all repopulated by progenitors of the previously non-growing phenotype in ampicillin-treated mice.

Amblypygids, colloquially known as whip spiders, are a charismatic order of arachnids that a characterised a unique pair of spined pedipalp appendages. Amblypygid pedipalps are hypothesised to primarily function as a prey capture device. However, the pedipalp is also used in several other functions including territorial contest and courtship, opening the possibility that the appendage could also be under the influence of sexual selection. There exists a vast degree of morphological diversity within the pedipalp, with relative length spanning nearly an order of magnitude across the group and spination varying markedly both within- and between species. The amblypygid pedipalp is therefore subject to multiple selective pressures, and both its external morphology and kinematics likely reflect this. Thus, the amblypygid pedipalp provides an ideal structure through which to study the evolution of morphological traits subject to multiple selective pressures, and the potential evolutionary trade-offs that may arise. Despite this, amblypygid pedipalp morphology and kinematics remains poorly quantified and little comparative work has been carried across the group. Here, I aim to quantify intra- and interspecific trends in amblypygid pedipalp shape and prey capture kinematics for the first time, using modern morphometric techniques and high-speed videography. In this work I present a comprehensive review of sexual dimorphism in arachnids and identify the common drivers behind this phenomenon. Building on this, I quantify sexual dimorphism in pedipalp size and shape in a single species of amblypygid using a novel geometric morphometric approach. This is followed by an broad analysis of intraspecific and interspecific trends in shape complexity of pedipalps using Elliptical Fourier Analysis. Finally, I draw a link between form and function by quantifying and comparing prey capture kinematics in a morphologically diverse set of amblypygids, using high-speed videography and motion analysis. This work provides new insights into amblypygid 8 pedipalp diversity and posits the possibility of an evolutionary trade-off between increased pedipalp length, for use in display in courtship and territorial contest, and strike performance during prey capture.