Search

Drug-target interactions (DTIs) prediction algorithms are used at various stages of the drug discovery process. In this context, specific problems such as deorphanization of a new therapeutic target or target identification of a drug candidate arising from phenotypic screens require large-scale predictions across the protein and molecule spaces. DTI prediction heavily relies on supervised learning algorithms that use known DTIs to learn associations between molecule and protein features, allowing for the prediction of new interactions based on learned patterns. The algorithms must be broadly applicable to enable reliable predictions, even in regions of the protein or molecule spaces where data may be scarce. In this paper, we address two key challenges to fulfill these goals: building large, high-quality training datasets and designing prediction methods that can scale, in order to be trained on such large datasets. First, we introduce LCIdb, a curated, large-sized dataset of DTIs, offering extensive coverage of both the molecule and druggable protein spaces. Notably, LCIdb contains a much higher number of molecules than publicly available benchmarks, expanding coverage of the molecule space. Second, we propose Komet (Kronecker Optimized METhod), a DTI prediction pipeline designed for scalability without compromising performance. Komet leverages a three-step framework, incorporating efficient computation choices tailored for large datasets and involving the Nyström approximation. Specifically, Komet employs a Kronecker interaction module for (molecule, protein) pairs, which efficiently captures determinants in DTIs, and whose structure allows for reduced computational complexity and quasi-Newton optimization, ensuring that the model can handle large training sets, without compromising on performance. Our method is implemented in open-source software, leveraging GPU parallel computation for efficiency. We demonstrate the interest of our pipeline on various datasets, showing that Komet displays superior scalability and prediction performance compared to state-of-the-art deep learning approaches. Additionally, we illustrate the generalization properties of Komet by showing its performance on an external dataset, and on the publicly available L H benchmark designed for scaffold hopping problems. Komet is available open source at https://komet.readthedocs.io and all datasets, including LCIdb, can be found at https://zenodo.org/records/10731712.

Inter-subject variability is a major hurdle for neuroimaging group-level inference, as it creates complex image patterns that are not captured by standard analysis models and jeopardizes the sensitivity of statistical procedures. A solution to this problem is to model random subjects effects by using the redundant information conveyed by multiple imaging contrasts. In this paper, we introduce a novel analysis framework, where we estimate the amount of variance that is fit by a random effects subspace learned on other images; we show that a principal component regression estimator outperforms other regression models and that it fits a significant proportion (10% to 25%) of the between-subject variability. This proves for the first time that the accumulation of contrasts in each individual can provide the basis for more sensitive neuroimaging group analyzes.

Functional Magnetic Resonance Imaging (fMRI) studies map the human brain by testing the response of groups of individuals to carefully-crafted and contrasted tasks in order to delineate specialized brain regions and networks. The number of functional networks extracted is limited by the number of subject-level contrasts and does not grow with the cohort. Here, we introduce a new group-level brain mapping strategy to differentiate many regions reflecting the variety of brain network configurations observed in the population. Based on the principle of functional segregation, our approach singles out functionally-specialized brain regions by learning group-level functional profiles on which the response of brain regions can be represented sparsely. We use a dictionary-learning formulation that can be solved efficiently with on-line algorithms, scaling to arbitrary large datasets. Importantly, we model inter-subject correspondence as structure imposed in the estimated functional profiles, integrating a structure-inducing regularization with no additional computational cost. On a large multi-subject study, our approach extracts a large number of brain networks with meaningful functional profiles.

Understanding brain structure and function entails the inclusion of anatomical and functional information in a common space, in order to study how these different informations relate to each other in a population of subjects. In this paper, we revisit the parcellation model and explicitly combine anatomical features, i.e. a segmentation of the cortex into gyri, with a functional information under the form of several cortical maps, which are used to further subdivide the gyri into functionally consistent regions. A probabilistic model is introduced, and the parcellation model is estimated using a Variational Bayes approach. The number of regions in the model is validated based on cross-validation. It is found that about 250 patches of cortex can be delineated both in the left and right hemisphere based on this procedure.

Inferring the position of functionally active regions from a multi-subject fMRI dataset involves the comparison of the individual data and the inference of a common activity model. While voxel-based analyzes, e.g. Random Effect statistics, are widely used, they do not model each individual activation pattern. Here, we develop a new procedure that extracts structures individually and compares them at the group level. For inference about spatial locations of interest, a Dirichlet Process Mixture Model is used. Finally, inter-subject correspondences are computed with Bayesian Network models. We show the power of the technique on both simulated and real datasets and compare it with standard inference techniques.

The analysis of functional magnetic resonance imaging (fMRI) data recorded on several subjects resorts to the so-called spatial normalization in a common reference space. This normalization is usually carried out on a voxel-by-voxel basis, assuming that after coregistration of the functional images with an anatomical template image in the Talairach reference system, a correct voxel-based inference can be carried out across subjects. Shortcomings of such approaches are often dealt with by spatially smoothing the data to increase the overlap between subject-specific activated regions. This procedure, however, cannot adapt to each anatomo-functional subject configuration. We introduce a novel technique for intra-subject parcellation based on spectral clustering that delineates homogeneous and connected regions. We also propose a hierarchical method to derive group parcels that are spatially coherent across subjects and functionally homogeneous. We show that we can obtain groups (or cliques) of parcels that well summarize inter-subject activations. We also show that the spatial relaxation embedded in our procedure improves the sensitivity of random-effect analysis.

Abstract: With the increasing number of available XML documents, numerous approaches for retrieval have been proposed in the literature. They usually use the tree representation of documents and queries to process them, whether in an implicit or explicit way. Although retrieving XML documents can be considered as a tree matching problem between the query tree and the document trees, only a few approaches take advantage of the algorithms and methods proposed by the graph theory. In this paper, we aim at studying the theoretical approaches proposed in the literature for tree matching and at seeing how these approaches have been adapted to XML querying and retrieval, from both an exact and an approximate matching perspective. This study will allow us to highlight theoretical aspects of graph theory that have not been yet explored in XML retrieval. [Copyright &y& Elsevier]