Abstract Motivation Manual curation of biomedical literature is slow and error-prone and while large language models trained on general texts have shown to be useful for text summarisation, these methods lack the domain-specific expertise required to perform this task accurately. Here we describe the creation of the first microbiome-specific text corpus, use this to train deep learning algorithms for named-entity recognition (NER) and entity linking (EL), and demonstrate their use to meta-analyse microbiome literature. Results The training and validation set (n = 1410) contained a total of 90 150 annotations (both long form and abbreviations). Using the gold-standard test set (n = 288), with an inter-annotator agreement rate of 99.52% for NER and 88.31% for EL, the trained models were evaluated and our fine-tuned BioBERT model achieved an F1-score of 96% for NER surpassing a rule- and dictionary-based annotation pipeline (94%). For EL the accuracy obtained by the deep learning models greatly surpassed that of the pipeline (91% vs 69%). Evaluated across the entire available literature (n = 6927) across 14 domains, our models annotate an entire full-text document in only 7 seconds. Availability All codes are available for automatic annotation and model training, with instructions on how to deploy the model on new text, from GitHub at https://github.com/omicsNLP/microbELP and Zenodo at https://dx.doi.org/10.5281/zenodo.20613467. The redistributable, annotated training set and unannotated test set are made available from Zenodo at https://dx.doi.org/10.5281/zenodo.17305410 with the redistributable, human-labelled test set hosted as benchmark on Codabench at https://www.codabench.org/competitions/10913/ (for NER only) and at https://www.codabench.org/competitions/11581/ (for NER+EL) for evaluation. The annotated documents for all available literature are hosted separately on Zenodo at https://dx.doi.org/10.5281/zenodo.17288826.

Minimally Invasive Surgery (MIS) reduces surgical risks and recovery times by enabling precise interventions around lesion sites. However, accessing these sites requires flexible instruments, which often compromise the precision afforded by rigid devices. This study proposes a thermally drawn fiber-based, tendon-driven soft robotic instrument with a two-stage hybrid motion control framework to enhance accuracy. Several learning-based inverse kinematic (IK) models were developed, with the LSTM-based model showing the best performance and used to guide open-loop large-scale motion, while a closed-loop controller refines accuracy via real-time feedback. The system is validated by path-following tasks and a simulated endometrial ablation in a vaginal phantom. Results show that the IK model realizes stable open-loop control with Euclidean error below 2 mm, while hybrid control further reduces errors to below 1 mm. This combination offers a promising MIS solution with high precision in difficult-to-reach surgical sites.

MOTIVATION: Random forests (RFs) can deal with a large number of variables, achieve reasonable prediction scores, and yield highly interpretable feature importance values. As such, RFs are appropriate models for feature selection and further dimension reduction. However, RFs are often not appropriate for correlated datasets due to their mode of selecting individual features for splitting. Addressing correlation relationships in high-dimensional datasets is imperative for reducing the number of variables that are assigned high importance, hence making the dimension reduction most efficient. Here, we propose the LAtent VAriable Stochastic Ensemble of Trees (LAVASET) method that derives latent variables based on the distance characteristics of each feature and aims to incorporate the correlation factor in the splitting step. RESULTS: Without compromising on performance in the majority of examples, LAVASET outperforms RF by accurately determining feature importance across all correlated variables and ensuring proper distribution of importance values. LAVASET yields mostly non-inferior prediction accuracies to traditional RFs when tested in simulated and real 1D datasets, as well as more complex and high-dimensional 3D datatypes. Unlike traditional RFs, LAVASET is unaffected by single 'important' noisy features (false positives), as it considers the local neighbourhood. LAVASET, therefore, highlights neighbourhoods of features, reflecting real signals that collectively impact the model's predictive ability. AVAILABILITY AND IMPLEMENTATION: LAVASET is freely available as a standalone package from https://github.com/melkasapi/LAVASET.

MOTIVATION: Data processing is a key bottleneck for 1H NMR-based metabolic profiling of complex biological mixtures, such as biofluids. These spectra typically contain several thousands of signals, corresponding to possibly few hundreds of metabolites. A number of binning-based methods have been proposed to reduce the dimensionality of 1 D 1H NMR datasets, including statistical recoupling of variables (SRV). Here, we introduce a new binning method, named JBA ("pJRES Binning Algorithm"), which aims to extend the applicability of SRV to pJRES spectra. RESULTS: The performance of JBA is comprehensively evaluated using 617 plasma 1H NMR spectra from the FGENTCARD cohort. The results presented here show that JBA exhibits higher sensitivity than SRV to detect peaks from low-abundance metabolites. In addition, JBA allows a more efficient removal of spectral variables corresponding to pure electronic noise, and this has a positive impact on multivariate model building. AVAILABILITY AND IMPLEMENTATION: The algorithm is implemented using the MWASTools R/Bioconductor package. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Summary: MWASTools is an R package designed to provide an integrated pipeline to analyse metabonomic data in large-scale epidemiological studies. Key functionalities of our package include: quality control analysis; metabolome-wide association analysis using various models (partial correlations, generalized linear models); visualization of statistical outcomes; metabolite assignment using statistical total correlation spectroscopy (STOCSY); and biological interpretation of metabolome-wide association studies results. Availability and implementation: The MWASTools R package is implemented in R (version > =3.4) and is available from Bioconductor: https://bioconductor.org/packages/MWASTools/. Contact: [email protected]. Supplementary information: Supplementary data are available at Bioinformatics online.

Summary: MetaboSignal is an R package that allows merging metabolic and signaling pathways reported in the Kyoto Encyclopaedia of Genes and Genomes (KEGG). It is a network-based approach designed to navigate through topological relationships between genes (signaling- or metabolic-genes) and metabolites, representing a powerful tool to investigate the genetic landscape of metabolic phenotypes. Availability and Implementation: MetaboSignal is available from Bioconductor: https://bioconductor.org/packages/MetaboSignal/. Contact: [email protected] . Supplementary information: Supplementary data are available at Bioinformatics online.

SUMMARY: MetaboNetworks is a tool to create custom sub-networks in Matlab using main reaction pairs as defined by the Kyoto Encyclopaedia of Genes and Genomes and can be used to explore transgenomic interactions, for example mammalian and bacterial associations. It calculates the shortest path between a set of metabolites (e.g. biomarkers from a metabonomic study) and plots the connectivity between metabolites as links in a network graph. The resulting graph can be edited and explored interactively. Furthermore, nodes and edges in the graph are linked to the Kyoto Encyclopaedia of Genes and Genomes compound and reaction pair web pages. AVAILABILITY AND IMPLEMENTATION: MetaboNetworks is available from http://www.mathworks.com/matlabcentral/fileexchange/42684. CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.