Biology subjects

Youssef, G.

Publications and source records attributed to Youssef, G..

11 recordsLinked to original sources

bioRxiv2026

Non-exonic read content reorganises lncRNA co-expression networks in renal and other cancers

Co-expression networks are used to find prognostic lncRNA programmes, but lncRNA quantification depends on unprocessed RNA in the library. In renal cancer (1,062 patients), the leading lncRNA axis tracked the non-exonic (non-feature) read fraction, a routine alignment metric, at Spearman rho 0.90. Without adjustment, most of the lncRNA network merged into one module. Adjustment preserved agreement with single-cell tumour-normal differences. Under rules fixed with the renal results known, 9 of 25 TCGA cancer types had a lncRNA-specific axis and 4 had network collapse (8 and 2 of 23 non-renal types). Intronic lncRNAs tracked the metric more than intergenic ones in all 25, and the metric was not consistently prognostic. All five renal modules that survived adjustment kept their direction in an external cohort, two nominally significant and none at a 5% false discovery rate. Three localised to malignant or epithelial cells. Together they added no discrimination over TNM and grade.

bioinformatics↗

bioRxiv2026

Malignant and stromal partial EMT programmes and cetuximab outcome in head and neck cancer

Partial epithelial-to-mesenchymal transition (pEMT) was defined in single malignant cells of head and neck squamous cell carcinoma but is usually scored in bulk tumours. In three single-cell datasets, its laminin-332 and hemidesmosome genes were expressed by malignant cells (the malignant arm) and its fibrillar collagens and vimentin by fibroblasts (the stromal arm). In bulk tumours the stromal arm dominated co-expression, so standard pEMT scores mainly measure stroma. We recovered the malignant arm with a classifier trained on single cells and a 12-gene core. In 40 patients treated with platinum and cetuximab, both malignant measures were higher in those with long progression-free survival (area under the curve 0.78 and 0.82), and the stromal arm was not. The malignant arm was inseparable from the Basal subtype, whose score did as well or better in these patients (0.93) and in two xenograft panels. Bulk pEMT scores should be read by cell of origin.

cancer biology↗

bioRxiv2026

Single-cell foundation models reveal context-sensitive cancer programmes under subtype shift

Single-cell foundation models (scFMs) have shown promise as transferable representations of cellular state, but recent zero-shot evaluations suggest that they do not consistently outperform simpler baselines. We asked whether this apparent limitation reflects an intrinsic weakness of scFMs or instead the difficulty of using them without task-specific adaptation. To test this, we fine-tuned two widely used scFMs, Geneformer and scGPT, on common tumour subtypes from renal, lung, and breast cancer, and compared them with a LightGBM baseline on within-domain validation cohorts and on out-of-domain rarer, unseen cancer subtypes. Across all three organs, the models achieved near-perfect within-domain discrimination (AUROC 0.98-1.00), but differences emerged under subtype shift. On chromophobe RCC, scGPT and Geneformer achieved AUROC 0.88 and 0.92 respectively versus 0.64 for LightGBM; on SCLC, Geneformer reached 1.00 versus 0.82 for LightGBM; and on TNBC, scGPT achieved 0.80 versus 0.49 for LightGBM. To determine whether this generalisation reflected meaningful adaptation rather than arbitrary feature drift, we applied Integrated Gradients, an interpretability technique, to the fine-tuned scFMs and SHAP to LightGBM. LightGBM showed highly stable gene-importance rankings across datasets, whereas the foundation models were substantially more context-sensitive. However, this flexibility was not random: all models converged on a shared within-domain core, while scFMs acquired larger rare-subtype-specific gene sets and pathway programmes during transfer. Pathway enrichment further supported the biological relevance of these attributed genes. Together, these results suggest that fine-tuned scFMs can bridge clinically relevant domain shifts in cancer single-cell analysis and that interpretability provides a practical route to distinguishing biologically grounded adaptation from rigid reuse of training-era rules.

bioinformatics↗

bioRxiv2025

A systems-level machine learning approach uncovers therapeutic targets in clear cell renal cell carcinoma

We present a generalisable, interpretable machine learning framework for therapeutic target discovery using single-cell transcriptomics, protein interaction networks, and drug proximity analysis. The pipeline integrates feature selection via gradient boosting classifiers, systems-level network inference, and in silico drug repurposing, enabling the identification of actionable targets with cellular specificity. As a proof of concept, we apply the method to clear cell renal cell carcinoma (ccRCC), an aggressive kidney cancer with limited treatment options. The model identifies 96 tumour-intrinsic genes, refines them to 16 targets through CRISPR screens and biological curation, and prioritises FDA-approved compounds via network-based proximity scoring. Several novel therapeutic mechanisms - including ABL1, CDK4/6, and JAK inhibition - emerge from this analysis, with predicted compounds showing superior efficacy to standard-of-care drugs across multiple ccRCC cell lines. Beyond ccRCC, this framework offers a scalable strategy for drug discovery across diverse diseases, combining machine learning interpretability with systems biology to accelerate therapeutic development.

systems biology↗

bioRxiv2024

Dissecting compounded hepatocyte injury in a model of MASLD progression from human induced pluripotent stem cells

Drug discovery for multifactorial diseases like metabolic dysfunction-associated steatotic liver disease (MASLD) remains challenging due to inadequate models and untargeted drug screenings. We combined stem-cell-based modeling with computational drug predictions identifying flavin pathways as therapeutic targets in MASLD. For disease stage-specific discovery, we established a MASLD testing model, compounding metabolic triggers to intensify mitochondrial crisis. In vitro injuries included adipo- and myokines, immune cell co-culture, and genomic risk factors. Benchmarking experiments revealed similarities with advanced human MASLD. To query therapeutic compounds, protein-protein-interaction networks, weighted gene co-expression, and knowledge graph-based analyses independently predicted flavin adenine dinucleotide (FAD) as an anti-MASLD factor. Dysregulated flavoproteomes in vitro and in vivo-in pediatric and adult MASLD patients- supported our flavin network-focused strategy. We established therapeutic FAD concentrations to mitigate metabolic injury and fibro-inflammation in human multicellular liver organoids and other assays. We enhanced therapeutic FAD effects through genetic mitochondrial biogenic augmentation and identified orally available flavo-active compounds--including Aspirin--restoring mitochondrial respiration. Our study demonstrates how integrating stem cell-derived disease modeling with computed drug predictions can expedite therapeutic discovery.

cell biology↗

bioRxiv2023

A CD24+CD271+ melanoma cancer stem cell generates a diffuse hierarchy of attributes that promote metastasis and therapeutic resistance

An important role for phenotype switching has been demonstrated in metastasis and therapeutic resistance of both melanoma and epithelial tumours. Phenotype switching in epithelial tumours is driven by a minority cancer stem cell sub-population with lineage plasticity, but such a sub-population has not been identified in melanoma. We investigated whether cell surface markers used to identify cancer stem cells in epithelial tumours could identify a cancer stem cell sub-population with lineage plasticity in melanoma. We identified a CD24+CD271+ minority sub-population in melanoma that possesses the stem cell characteristics of lineage plasticity and self-renewal. This population displayed hybrid characteristics, combining the attributes of discrete CD24+ and CD271+ cellular sub-populations but with heightened sphere formation, lineage plasticity, migratory ability and drug resistance over its single-marker counterparts. CD24+ and CD24+CD271+ stem cell sub-populations were observed in 10% of human melanomas, mainly at the invasive front. The lack of CD24+ and CD24+CD271+ stem cells in the majority of human melanoma specimens led us to conclude that they may be dispensable for melanoma progression. Nevertheless, the enhanced sphere formation, lineage plasticity, migratory ability and drug resistance of the CD24+CD271+ sub-population may signal a contextual requirement for these stem cells when melanomas face challenging environments both clinically and in experimental systems.

cancer biology↗

bioRxiv2022

Chronic activation of Toll-like receptor 2 induces an ichthyotic skin phenotype

Ichthyosis defines a group of chronic conditions that manifest phenotypically as a thick layer of fish-like scales in response to disorders of cornification and often affects the entire skin. While the gene mutations that lead to ichthyosis are well documented, the actual signalling mechanisms that lead to scaling are poorly characterised, however recent publications suggest that there are common mechanisms active in ichthyotic tissue, and in analogous models of ichthyosis. Combining gene expression analysis of gene-specific shRNA knockdowns of more severe autosomal recessive congenital ichthyoses (ARCI) and proteomic analysis of skin scale from ARCI patients, we identified a common activation of the Toll-like receptor (TLR) 2 pathway. Exogenous activation of TLR2 led to increased expression of important cornified envelope genes and in organotypic culture caused hyperkeratosis. Conversely blockade of TLR2 signalling in ichthyosis patient keratinocytes and our shRNA models reduced the expression of keratin 1, a structural protein overexpressed in ichthyosis scale. A time-course of Tlr2 activation in rat epidermal keratinocytes revealed that although there was rapid initial activation of innate immune pathways, this was rapidly superseded by widespread up-regulation of epidermal differentiation related proteins. Both NF{kappa}{beta} phosphorylation and the Gata3 up-regulation was associated with this switch and Gata3 overexpression was sufficient to increase Keratin 1 expression. Taken together, these data define a dual role for Tlr2 during epidermal barrier repair, that may be a useful therapeutic modality in treating diseases of epidermal barrier dysfunction.

cell biology↗

bioRxiv2022

dialogi: Utilising NLP with chemical and disease similarities to drive the identification of Drug-Induced Liver Injury literature

Drug-Induced Liver Injury (DILI), despite its low occurrence rate, can cause severe side effects or even lead to death. Thus, it is one of the leading causes for terminating the development of new, and restricting the use of already-circulating, drugs. Moreover, its multifactorial nature, combined with a clinical presentation that often mimics other liver diseases, complicate the identification of DILI-related literature, which remains the main medium for sourcing results from the clinical practice and experimental studies. In this work- contributing to the Literature AI for DILI Challenge of the Critical Assessment of Massive Data Analysis (CAMDA) 2021- we present an automated pipeline for distinguishing between DILI-positive and negative papers. We used Natural Language Processing (NLP) to filter out the uninformative parts of a text, and identify and extract mentions of chemicals and diseases. We combined that information with small-molecule and disease embeddings, which are capable of capturing chemical and disease similarities, to improve classification performance. The former are directly sourced from the Chemical Checker (CC). For the latter, we collected data that encode different aspects of disease similarity from the National Library of Medicines (NLM) Medical Subject Headings (MeSH) thesaurus and the Comparative Toxicogenomics Database (CTD). Following a similar procedure as the one used in the CC, vector representations for diseases were learnt and evaluated. Two Neural Network (NN) classifiers were developed: one that only accepts texts as input (baseline model) and an augmented classifier that also utilises chemical and disease embeddings (extended model). We trained, validated, and tested the models through a Nested Cross-Validation (NCV) scheme with 10 outer and 5 inner folds. During this, the baseline and extended models performed virtually identically, with macro F1-scores of 95.04 {+/-} 0.61% and 94.80 {+/-} 0.41%, respectively. Upon validation on an external, withheld, dataset, representing imbalanced data, the extended model achieved an F1-score of 91.14 {+/-} 1.62%, outperforming its baseline counterpart, which got a lower score of 88.30 {+/-} 2.44%. We make further comparisons between the classifiers and discuss future improvements and directions, including utilising chemical and disease embeddings for visualisation and exploratory analysis of the DILI-positive literature.

bioinformatics↗

bioRxiv2022

DILIc: An AI-based classifier to search for Drug-Induced Liver Injury literature

Drug-Induced Liver Injury (DILI) is a class of Adverse Drug Reactions (ADR) which causes problems in both clinical and research settings. It is the most frequent cause of acute liver failure in the majority of western countries and is a major cause of attrition of novel drug candidates. Manual trawling of literature for is the main route of deriving information on DILI from research studies. This makes it an inefficient process prone to human error. Therefore, an automatized AI model capable of retrieving DILI-related papers from the huge ocean of literature could be invaluable for the drug discovery community. In this project, we built an artificial intelligence (AI) model combining the power of Natural Language Processing (NLP) and Machine Learning (ML) to address this problem. This model uses NLP to filter out meaningless text (e.g. stopwords) and uses customized functions to extract relevant keywords as singleton, pair, triplet and so on. These keywords are processed by apriori pattern mining algorithm to extract relevant patterns which are used to estimate initial weightings for a ML classifier. Along with pattern importance and frequency, an FDA-approved drug list mentioning DILI adds extra confidence in classification. The combined power of these methods build a DILI classifier (DILIC) with 94.91% cross-validation and 94.14% external validation accuracy. To make DILIC as accessible as possible, including to researchers without coding experience, an R Shiny App capable of classifing single or multiple entries for DILI is developed to enhance ease of user experience and made available at https://researchmind.co.uk/diliclassifier/).

bioinformatics↗

bioRxiv2020

A longitudinal analysis of puberty-related cortical development

The brain undergoes extensive structural changes during adolescence, concurrent to puberty-related physical and hormonal changes. While animal research suggests these biological processes are related to one another, our knowledge of brain development in humans is largely based on age-related processes. Thus, the current study characterized puberty-related changes in human brain structure, by combining data from two longitudinal neuroimaging cohorts. Beyond normative changes in cortical thickness, we examined whether individual differences in the rate of pubertal maturation (or "pubertal tempo") was associated with variations in cortical trajectories. Participants (N = 192; scans = 366) completed up to three waves of MRI assessments between 8.5 and 14.5 years of age, as well as questionnaire assessments of pubertal stage at each wave. Generalized additive mixture models were used to characterize trajectories of cortical development. Results revealed widespread linear puberty-related changes across much of the cortex. Many of these changes, particularly within the frontal and parietal cortices, were independent of age-related development. Males exhibiting faster pubertal tempo demonstrated greater thinning in the precuneus and frontal cortices than same-aged and -sex peers. Findings suggest that the unique influence of puberty on cortical development may be more extensive than previously identified, and also emphasize important individual differences in the coupling of these developmental processes.

neuroscience↗

bioRxiv2020

Disseminating cells in human tumours acquire an EMT stem cell state that is predictive of metastasis

Cancer stem cells (CSCs) undergo epithelial-mesenchymal transition (EMT) to drive metastatic dissemination in experimental cancer models. However, tumour cells undergoing EMT have not been observed disseminating into the tissue surrounding human tumour specimens, leaving the relevance to human cancer uncertain. We have previously identified both EpCAM and CD24 as markers of EMT CSCs with enhanced plasticity. This afforded the opportunity to investigate whether retention of EpCAM and CD24 alongside upregulation of the EMT marker Vimentin can identify disseminating EMT CSCs in human tumours. Examining disseminating tumour cells in over 12,000 imaging fields from 84 human oral cancer specimens, we see a significant enrichment of single EpCAM, CD24 and Vimentin co-stained cells disseminating beyond the tumour body in metastatic specimens. Through training an artificial neural network, these predict metastasis with high accuracy (cross-validated accuracy of 87-89%). In this study, we have observed single disseminating EMT CSCs in human oral cancer specimens, and these are highly predictive of metastatic disease.

cancer biology↗