Biology subjects

Soul, J.

Publications and source records attributed to Soul, J..

6 recordsLinked to original sources

bioRxiv2026

A low-annotation-budget PubMedBERT classifier for chondrogenesis regulator discovery via active learning

Motivation: Biomedical natural language processing (Bio-NLP) classification tasks are often limited by the cost of manual annotation, especially for specialised extraction problems where labelled corpora are scarce. Active learning can reduce this cost by prioritising informative samples, while large language models (LLMs) avoid task-specific annotation, but at high computational cost. We ask whether an encoder trained via active learning can match LLM performance at a lower computational cost, using identification of chondrogenesis regulators from PubMed abstracts as a test case applicable to other under-annotated Bio-NLP problems. Results: We built an active learning framework to fine-tune PubMedBERT to classify gene mentions as regulators of chondrocyte differentiation, achieving an AUC-ROC of 0.93 using 1,238 annotated sentences. Against open-weight LLMs (Qwen3, Llama-3.1), PubMedBERT matched the top-performing LLMs on AUC-ROC and achieved the highest precision (0.69) of any model, while classifying the full corpus far faster. Applied to 67,111 gene mentions, it identified 1,128 candidate regulators, recovered 79% of Gene Ontology-annotated genes related to chondrocyte differentiation (94 genes recovered), and proposed 981 new candidates. The pipeline is agnostic to the target process, suggesting it could be adapted to identify regulators of other biological processes. Availability and Implementation: Code is available at https://github.com/ChondroTextomics/ALRegulatorDiscovery. Data are available at Zenodo: https://doi.org/10.5281/zenodo.22746090. Final model is available in HuggingFace Hub amav/pubmedbert-chondrogenesis-classifier Supplementary information: Supplementary data are available in a separate file

bioinformatics↗

bioRxiv2026

Automated generation of a gene perturbation transcriptomic atlas using large language models

Public transcriptomic repositories contain thousands of gene perturbation experiments, a valuable resource for understanding gene function, but perturbation metadata are not structured, which blocks systematic reuse. Existing perturbation atlases depend on expert manual curation, so they are costly to maintain and infrequently updated, while automated grouping approaches neither identify which samples form the perturbation arm nor recover the perturbed gene. Here we develop an automated pipeline that uses large language models to find single-gene perturbation experiments in NCBI-GEO and reconstruct their case-control sample groupings, along with the perturbed gene, perturbation type and cell line as structured, ontology-normalised fields. We manually curated 3,300 GEO experiments with sample-level case-control assignments and release these as an open benchmark (2,400 training, 600 validation, 300 temporally held-out test). Reasoning models and task-specific finetuning substantially improved identification of valid perturbation groups, with the best model reaching precision 0.925 and recall 0.836 on the test set. Applied at scale, the pipeline generated an atlas of 6,802 gene perturbation expression signatures from 4,453 GEO experiments, covering 2,907 uniquely perturbed genes. An R package, perturbMatch, supports exploration of the atlas and querying of user-supplied expression signatures against it using similarity scoring, so users can identify experiments that recapitulate a transcriptional state of interest.

bioinformatics↗

bioRxiv2023

miR-324 mediates bone homeostasis through the regulation of osteoblast and osteoclast differentiation and activity

microRNAs (miRNAs) are non-coding RNAs which modulate the expression of other RNA molecules. One miRNA can target many transcripts, allowing each miRNA to play key roles in many biological pathways. miR-324 is a miRNA previously implicated in bone and cartilage maintenance, defects of which result in common age-related diseases, such as osteoporosis or osteoarthritis (OA). In global miR-324-null mice cartilage damage was increased in both surgically and ageing-induced OA, despite minimal changes to the cartilage transcriptome, with few predicted miR-324 targets dysregulated. However, micro-computed tomography and histology demonstrated that global miR- 324-null the mice had an increase in bone mineral density, trabecular thickness and cortical thickness, with many parameters increasing with age. The bone marrow of miR-324-null mice also had reduced lipid content while and in vivo TRAP staining revealed a decrease in osteoclasts, with histomorphometry demonstrating an increased rate of bone formation in miR-324-null mice. Ex vivo assays revealed that the high bone mass phenotype of the miR-324-null mice resulted from increased osteoblast activity and decreased osteoclastogenesis. RNA-seq and qRT-PCR followed by miR-324 target prediction and validation in osteoblasts, osteoclasts and bone marrow macrophages identified the osteoclast fusion regulator Pin1 as a miR-324 target in the osteoclast lineage and the master osteogenic regulator Runx2 as a target of miR-324-5p in osteoblasts, the in vitro overexpression of which recapitulated the increased osteogenesis and decreased adipogenesis phenotype observed in vivo. These data point to important roles of miR-324 in skeletal biology with altered bone homeostasis in miR-324-null mice potentially causal for the increased cartilage damage observed during OA and ageing. Elucidation of pathways regulated by miR-324 offer promise for the treatment of bone diseases such as osteoporosis.

cell biology↗

bioRxiv2023

Systems analysis of miR-199a/b-5p and multiple miR-199a/b-5p targets during chondrogenesis

Changes in chondrocyte gene expression can contribute to the development of osteoarthritis (OA), and so recognition of the regulative processes during chondrogenesis can lead to a better understanding of OA. microRNAs (miRNAs) are key regulators of gene expression in chondrocytes/OA and we have used a combined experimental, bioinformatic, and systems biology approach to explore the multiple miRNA-mRNA interactions that regulate chondrogenesis. A longitudinal chondrogenesis bioinformatic analysis identified paralogues miR-199a-5p and miR-199b-5p as pro-chondrogenic regulators. Experimental work demonstrated alteration of miR-199a-5p or miR-199b-5p expression led to significant inverse modulation of key chondrogenic genes and extracellular matrix production. miR-199a/b-5p targets FZD6, ITGA3 and CAV1 were identified by inhibition experiments and verified as direct targets by luciferase assay. The experimental work was used to generate and parameterize a multi-miRNA 14-day chondrogenesis kinetic model to be used as a repository for the experimental work and as a resource for further investigation of this system. This is the first multi-miRNA model of a chondrogenesis-based system, and highlights the complex relationships between regulatory miRNAs, and their target mRNAs.

systems biology↗

bioRxiv2023

Mesenchymal Stromal Cell secretome is affected by tissue source, donor age and sex.

Variation in Mesenchymal Stromal Cell (MSC) function depending on their origin is problematic, as it may confound clinical outcomes of MSC therapy. Current evidence suggests that the therapeutic benefits of MSCs is primarily attributed to secretion of various biologically active factors (secretome). However, the effect of donor characteristics on the MSC secretome composition remains largely unknown. Here, we examined the influence of donor age, sex and tissue source, on the protein profile of the equine MSC secretome. Initially, we used dynamic metabolic labelling with stable isotopes combined with liquid chromatography-tandem mass spectrometry (LC-MS/MS) to identify secreted proteins in MSC conditioned media (CM). Seventy proteins were classified as classically-secreted based on the rate of isotope label incorporation into newly synthesised proteins released into the extracellular space. Next, we analysed CM of bone marrow- (n = 14) and adipose-derived MSCs (n = 16) with label-free LC-MS/MS proteomics. Clustering analysis of 314 proteins detected across all samples identified tissue source as the main factor driving variability in MSC CM proteomes. Linear modelling applied to the subset of 70 secreted proteins identified tissue-related difference in the abundance of 23 proteins. There was an age-related decrease in the abundance of two proteins (CTHRC1, LOX), which has been validated with western blot and enzymatic activity assay. There was limited evidence of sex-related differences in protein abundance. In conclusion, this study provides evidence that tissue source and donor age contribute to heterogeneity in the protein composition of MSC secretomes which may influence the effects of MSC-based cell therapy. Significance statementThis research shows for the first time that donor age can influence the proteins secreted by mesenchymal stromal cells (MSC), which is considered the main mechanism through which they mediate their biological actions. This information may improve our understanding of the variable outcomes observed in MSC clinical studies, as well as identify optimal sources and donor selection for specific clinical applications.

cell biology↗

bioRxiv2020

microRNA-seq of cartilage reveals an over-abundance of miR-140-3p which contains functional isomiRs

MiR-140 is selectively expressed in cartilage. Deletion of the entire miR-140 locus in mice results in growth retardation and early-onset osteoarthritis-like pathology, however the relative contribution of miR-140-5p or miR-140-3p to the phenotype remains to be determined. An unbiased small RNA sequencing approach identified miR-140-3p as significantly more abundant (>10-fold) than miR-140-5p in human cartilage. Analysis of these data identified multiple miR-140-3p isomiRs differing from the miRBase annotation at both the 5 and 3 end, with >99% having one of two seed sequences (5 bases 2-8). Canonical (miR-140-3p.2) and shifted (miR-140-3p.1) seed isomiRs were overexpressed in chondrocytes and transcriptomics performed to identify targets. miR-140-3p.1 and miR-140-3p.2 significantly down-regulated 694 and 238 genes respectively, of which only 162 genes were commonly down-regulated. IsomiR targets were validated using 3 UTR luciferase assays. miR-140-3p.1 targets were enriched within up-regulated genes in rib chondrocytes of Mir140-null mice and within down-regulated genes during human chondrogenesis. Finally, through imputing the expression of miR-140 from the expression of the host gene WWP2 in 124 previously published datasets, an inverse correlation with miR-140-3p.1 predicted targets was identified. Together these data suggest the novel seed containing isomiR miR-140-3p.1 is more functional than original consensus miR-140-3p seed containing isomiR.

molecular biology↗