Predicting tissue-specific protein functions and proteinprotein interactions (PPI) is essential for understanding human biology, diseases, and potential therapeutics.Recently, as a promising direction, more and more complex unsupervised feature learning approaches have emerged in the field, but none of them consider the scale-free nature and the underlying geometry of multi-layer PPI networks.Therefore, this study proposes contextualized, tissue-specific representation learning in non-Euclidean geometries and demonstrates that hyperbolic embeddings capture the structure of multi-layer PPI networks with less distortion and achieve better performance in tissue-specific protein function prediction.

This paper investigates payload grasping from a moving platform using a hook-equipped aerial manipulator. First, a computationally efficient trajectory optimization based on complementarity constraints is proposed to determine the optimal grasping time. To enable application in complex, dynamically changing environments, the future motion of the payload is predicted using a physics simulator-based model. The success of payload grasping under model uncertainties and external disturbances is formally verified through a robustness analysis method based on integral quadratic constraints. The proposed algorithms are evaluated in a high-fidelity physical simulator, and in real flight experiments using a customdesigned aerial manipulator platform.

BACKGROUND: Gene regulatory network modeling is a complex structure learning problem that involves both observational data analysis and experimental interventions. Bayesian causal discovery provides a principled framework for modeling observational data, generating posterior distributions that best represent the underlying structure. While recent algorithms offer efficient and accurate structure learning, integrating experiment design can further enhance predictive performance. RESULTS: We introduce novel acquisition functions for experiment design in gene expression data, leveraging active learning in both Essential Graph and Graphical Model spaces. We evaluate scalable structure learning algorithms within an active learning framework to optimize intervention selection. Our study explores existing active learning strategies, adapts techniques from other domains to structure learning, and proposes a novel approach using Equivalence Class Entropy Sampling (ECES) and Equivalence Class BALD Sampling (EBALD). Using DREAM4's Gene Net Weaver and Sachs protein signaling data, we assess the effectiveness of different strategies in improving network learning. CONCLUSION: Existing Bayesian experiment design strategies often overlook the Essential Graph structure, making inference more challenging due to the large number of possible graphs. Our results demonstrate that integrating active learning into structure learning algorithms can significantly improve performance, offering a scalable and effective approach for gene regulatory network discovery.

Differentially expressed gene (DEG) signatures play a crucial role in transcriptomics, offering insight into cellular responses and disease mechanisms, thus accelerating drug and target discovery. Understanding the geometric structure of the DEG space is essential for developing more effective computational methods. In this study, we show that the DEG signature space exhibits a scale-free nature, indicating an underlying hyperbolic geometry. To demonstrate the practical implications of this finding, we conducted a comparative analysis of unsupervised dimensionality reduction techniques, evaluating them based on local and global structure preservation. Our results indicate that hyperbolic embeddings better capture DEG signatures, supporting our claim on their underlying geometry. Besides, our prior results on drug-target interaction prediction suggest that hyperbolic embeddings improve performance when DEG signatures are used as input, reinforcing their effectiveness in downstream supervised predictive tasks as well. These findings highlight the relevance of hyperbolic geometry in modeling DEG signatures, suggesting future directions for machine learning applications in transcriptomics.

In biomarker research, there is a growing demand for computational methods to efficiently identify novel metabolite-disease associations (MDAs).Current approaches, however, do not take into account the underlying geometry of the MDA space.Here, we show that classifiers leveraging hyperbolic embeddings achieve comparable results to their Euclidean counterparts with significantly lower dimensionality, aligning better with the association network's scale-free nature.Finally, through a case study, we provide an interpretation of the model embeddings and investigate newly predicted associations.Our results demonstrate the intrinsic non-Euclidean geometry of the MDA space, providing direction for further research.A Pytorch-based implementation is available at https://github.com/PDomonkos/hyperbolic-MDA-prediction.

MOTIVATION: Identifying disease modules within molecular interaction networks is an essential exploratory step in computational biology, offering insights into disease mechanisms and potential therapeutic targets. Traditional methods often struggle with the inherent complexity and overlapping nature of biological networks, and they are limited in effectively leveraging the vast amount of available genomic data and biological knowledge. This limitation underscores the need for more effective, automated approaches to integrate these rich data sources. RESULTS: In this work, we propose GNN4DM, a novel graph neural network-based structured model that automates the discovery of overlapping functional disease modules. GNN4DM effectively integrates network topology with genomic data to learn the representations of the genes corresponding to functional modules and align these with known biological pathways for enhanced interpretability. Following the DREAM benchmark evaluation setting and extending with three independent data sources (GWAS Atlas, FinnGen, and DisGeNET), we show that GNN4DM performs better than several state-of-the-art methods in detecting biologically meaningful modules. Moreover, we demonstrate the method's applicability by discovering two novel multimorbidity modules significantly enriched across a diverse range of seemingly unrelated diseases. AVAILABILITY AND IMPLEMENTATION: Source code, all training data, and all identified disease modules are freely available for download at https://github.com/gezsi/gnn4dm. GNN4DM was implemented in Python.

To apply federated learning to drug discovery we developed a novel platform in the context of European Innovative Medicines Initiative (IMI) project MELLODDY (grant n°831472), which was comprised of 10 pharmaceutical companies, academic research labs, large industrial companies and startups. The MELLODDY platform was the first industry-scale platform to enable the creation of a global federated model for drug discovery without sharing the confidential data sets of the individual partners. The federated model was trained on the platform by aggregating the gradients of all contributing partners in a cryptographic, secure way following each training iteration. The platform was deployed on an Amazon Web Services (AWS) multi-account architecture running Kubernetes clusters in private subnets. Organisationally, the roles of the different partners were codified as different rights and permissions on the platform and administrated in a decentralized way. The MELLODDY platform generated new scientific discoveries which are described in a companion paper.

High dimensional genetic data is widely used to explore simple relations between traits, diseases and relevant genetic factors with high effect size, although identifying factors with small effect sizes or synergistic effects needs sophisticated solutions. A promising option is the application of network propagation methods for amplifying genome-wide association (GWA) results by incorporating a large amount of biological knowledge. This approach is also supported by the increasing availability of GWA summary statistics for comprehensive sets of phenotypes, frequently from multiple biobanks. However, the application of network propagation methods in GWA context is still in its early phase, despite its established role in the analysis of gene expression data and rare variants. First, we introduce the complete network-based GWAS workflow, also extending it for the simultaneous analysis of multiple traits and diseases. Second, we overview critical steps, possible solutions, and publicly available resources for this workflow; namely (1) the reliability of GWA results, (2) gene definitions and aggregation methods, (3) context-specific molecular networks, and (4) network propagation methods. Third, we present results from our large-scale evaluation of these options, such as established gene-disease relations, reference pathways, and recent public GWA results for hundreds of phenotypes from the UK Biobank. Results show serious inconsistencies in all settings regarding tested phenotypes, input transformations, networks, and network propagation method. This suggests a central unresolved issue in the application of this methodology for amplifying GWA results, and we hypothesize that the currently applied molecular networks form a serious bottleneck for the much expected multi-trait analysis.

Uncertainty over model structures poses a challenge for many approaches exploring effect strength parameters at system-level. Monte Carlo methods for full Bayesian model averaging over model structures require considerable computational resources, whereas bootstrapped graphical lasso and its approximations offer scalable alternatives with lower complexity. Although the computational efficiency of graphical lasso based approaches has prompted growing number of applications, the restrictive assumptions of this approach are frequently ignored. We demonstrate using an artificial and a real-world example that full Bayesian averaging using Bayesian networks provides detailed estimates through posterior distributions for structural and parametric uncertainties and it is a feasible alternative, which is routinely applicable in mid-sized biomedical problems with hundreds of variables. We compare Bayesian estimates with corresponding frequentist quantities from bootstrapped graphical lasso using pairwise Markov Random Fields, discussing also their different interpretations. We present results using synthetic data from an artificial model and using the UK Biobank data set to construct a psychopathological network centered around depression (this research has been conducted using the UK Biobank Resource under Application Number 1602).

BACKGROUND: Computational fusion approaches to drug-target interaction (DTI) prediction, capable of utilizing multiple sources of background knowledge, were reported to achieve superior predictive performance in multiple studies. Other studies showed that specificities of the DTI task, such as weighting the observations and focusing the side information are also vital for reaching top performance. METHOD: We present Variational Bayesian Multiple Kernel Logistic Matrix Factorization (VB-MK-LMF), which unifies the advantages of (1) multiple kernel learning, (2) weighted observations, (3) graph Laplacian regularization, and (4) explicit modeling of probabilities of binary drug-target interactions. RESULTS: VB-MK-LMF achieves significantly better predictive performance in standard benchmarks compared to state-of-the-art methods, which can be traced back to multiple factors. The systematic evaluation of the effect of multiple kernels confirm their benefits, but also highlights the limitations of linear kernel combinations, already recognized in other fields. The analysis of the effect of prior kernels using varying sample sizes sheds light on the balance of data and knowledge in DTI tasks and on the rate at which the effect of priors vanishes. This also shows the existence of "small sample size" regions where using side information offers significant gains. Alongside favorable predictive performance, a notable property of MF methods is that they provide a unified space for drugs and targets using latent representations. Compared to earlier studies, the dimensionality of this space proved to be surprisingly low, which makes the latent representations constructed by VB-ML-LMF especially well-suited for visual analytics. The probabilistic nature of the predictions allows the calculation of the expected values of hits in functionally relevant sets, which we demonstrate by predicting drug promiscuity. The variational Bayesian approximation is also implemented for general purpose graphics processing units yielding significantly improved computational time. CONCLUSION: In standard benchmarks, VB-MK-LMF shows significantly improved predictive performance in a wide range of settings. Beyond these benchmarks, another contribution of our work is highlighting and providing estimates for further pharmaceutically relevant quantities, such as promiscuity, druggability and total number of interactions.

Comorbidity patterns have become a major source of information to explore shared mechanisms of pathogenesis between disorders. In hypothesis-free exploration of comorbid conditions, disease-disease networks are usually identified by pairwise methods. However, interpretation of the results is hindered by several confounders. In particular a very large number of pairwise associations can arise indirectly through other comorbidity associations and they increase exponentially with the increasing breadth of the investigated diseases. To investigate and filter this effect, we computed and compared pairwise approaches with a systems-based method, which constructs a sparse Bayesian direct multimorbidity map (BDMM) by systematically eliminating disease-mediated comorbidity relations. Additionally, focusing on depression-related parts of the BDMM, we evaluated correspondence with results from logistic regression, text-mining and molecular-level measures for comorbidities such as genetic overlap and the interactome-based association score. We used a subset of the UK Biobank Resource, a cross-sectional dataset including 247 diseases and 117,392 participants who filled out a detailed questionnaire about mental health. The sparse comorbidity map confirmed that depressed patients frequently suffer from both psychiatric and somatic comorbid disorders. Notably, anxiety and obesity show strong and direct relationships with depression. The BDMM identified further directly co-morbid somatic disorders, e.g. irritable bowel syndrome, fibromyalgia, or migraine. Using the subnetwork of depression and metabolic disorders for functional analysis, the interactome-based system-level score showed the best agreement with the sparse disease network. This indicates that these epidemiologically strong disease-disease relations have improved correspondence with expected molecular-level mechanisms. The substantially fewer number of comorbidity relations in the BDMM compared to pairwise methods implies that biologically meaningful comorbid relations may be less frequent than earlier pairwise methods suggested. The computed interactive comprehensive multimorbidity views over the diseasome are available on the web at Co=MorNet: bioinformatics.mit.bme.hu/UKBNetworks.

We have developed a World Wide Web application for the collection of EPRs (electronic patient records) from uterine adnexal masses pre-operatively examined with transvaginal ultrasonography. The application has been used intensively since November 2000 by nine of the 19 international centers that joined the International Ovarian Tumor Analysis (IOTA) consortium. The IOTA database contains 68 parameters for 1,150 masses. We report the design and implementation of the generic Web-based clinical data entry system and describe the advantages and drawbacks that we have experienced while developing, using and maintaining the system. The data model, the user interface, the help system, the constraints (mandatory/optional) and the quality checking were all based on the medical protocol created by the IOTA consortium. The data collection system has become an open and transparent implementation of the formalized protocol. It covers the complete path of the patient data from the clinical situation to the finalized database. This approach provides new types of possibilities for the data analysis, since all aspects of the data collection are documented and formally available to the data analyst. The IOTA Web site can be found at, which also serves as the entry point for the secure EPR application.

The increasing amount and variety of domain knowledge and the availability of increasingly large quantities of electronic literature requires new types of support for the development of complex knowledge models. P. Antal et al. (2001) proposed the application of so-called annotated Bayesian networks (ABNs), which are textually-enriched probabilistic domain models that help knowledge engineers and medical experts to find and organize the information that is necessary in model-building. In this paper, we describe an information retrieval language in which the formalized domain knowledge and the attached textual information can be accessed in an integrated fashion and can be used to define various retrieval schemes and relevance measures. This language on the one hand provides maximum flexibility for knowledge engineers to exploit the available annotated domain model as contextual information. On the other hand, it allows the definition of complex, high-level queries, in which the contextual use of the annotated domain model can be optimized for clinical situations. We compare the performance of the standard and the proposed query language in the ovarian cancer domain.

Thanks to its increasing availability, electronic literature can now be a major source of information when developing complex statistical models where data is scarce or contains much noise. This raises the question of how to integrate information from domain literature with statistical data. Because quantifying similarities or dependencies between variables is a basic building block in knowledge discovery, we consider here the following question. Which vector representations of text and which statistical scores of similarity or dependency support best the use of literature in statistical models? For the text source, we assume to have annotations for the domain variables as short free-text descriptions and optionally to have a large literature repository from which we can further expand the annotations. For evaluation, we contrast the variables similarities or dependencies obtained from text using different annotation sources and vector representations with those obtained from measurement data or expert assessments. Specifically, we consider two learning problems: clustering and Bayesian network learning. Firstly, we report performance (against an expert reference) for clustering yeast genes from textual annotations. Secondly, we assess the agreement between text-based and data-based scores of variable dependencies when learning Bayesian network substructures for the task of modeling the joint distribution of clinical measurements of ovarian tumors.

We have previously (2000) reported on the development of Bayesian network models for the pre-operative discrimination between malignant and benign ovarian masses. The models incorporated both medical background knowledge and patient data, which required the traceability of the incorporated prior medical knowledge. For this purpose, we followed a particular annotation method for Bayesian networks using a dedicated representation. In this paper, we present the resulting annotated Bayesian network (ABN) representation that consists of a regular Bayesian network, with standard probabilistic semantics, and a corresponding semantic network, to which textual information sources are attached. We demonstrate the applicability of such a dual model to represent both the rigorous probabilistic and the unconstrained textual medical knowledge. We describe methods on how these ABN models can be used: (1) as a domain model to arrange the personal textual information of a clinician according to the semantics of the domain, (2) in decision support to provide detailed (and even personalized) explanation, and (3) to enhance the information retrieval to find new textual information more efficiently.

The pre-operative discrimination between malignant and benign masses is a crucial issue in gynaecology. Next to the large amount of background knowledge, there is a growing amount of collected patient data that can be used in inductive techniques. These two sources of information result in two different modelling strategies. Based on the background knowledge, various discrimination models have been constructed by leading experts in the field, tuned and tested by observations. Based on the patient observations, various statistical models have been developed, such as logistic regression models and artificial neural network models. For the efficient combination of prior background knowledge and observations, Bayesian network models are suggested. We summarize the applicability of this technique, report the performance of such models in ovarian cancer diagnosis and outline a possible hybrid usage of this technique.