Biology subjects

Amini, P.

Publications and source records attributed to Amini, P..

6 recordsLinked to original sources

bioRxiv2025

Stability and Performance of Linear Combination Tests of Gene Set Enrichment for Multiple Covariance Estimators in Unbalanced Studies

Gene set analysis (GSA) is essential for understanding coordinated gene expression changes within biological pathways, especially in high-dimensional data generated by platforms such as RNA-seq and microarrays. This study focuses on the linear combination test (LCT), a GSA method that combines multiple genelevel statistics into a powerful test statistic to assess the association between a gene set and outcomes of interest in a given set of samples. We evaluated the performance and stability of LCT using different covariance matrix estimators, including ridge, graphical lasso, and adaptive lasso, which are known for their effectiveness in high-dimensional data analysis. In addition, we assessed the robustness of LCT in the face of unbalanced study designs, which are typical in biomedical research due to limited sample availability and the high cost of data generation. We conducted a simulation study and applied LCT to publicly available gene expression datasets comparing patients with systemic lupus erythematosus (SLE) to healthy controls, where the number of controls is significantly lower than the number of cases. Our findings demonstrate that while LCTs default shrinkage estimator shows limitations in highly correlated and unbalanced designs, ridge estimation provides a more reliable alternative for unbalanced scenarios. Researchers can optimize LCTs performance by selecting appropriate covariance estimators based on their data structure. These results suggest that LCT is a reliable and powerful tool for GSA in unbalanced studies, identifying SLE-relevant gene sets more effectively than other GSA methods and showing validation against clinical phenotypes, offering valuable insights into the underlying mechanisms of complex diseases such as SLE.

bioinformatics↗

bioRxiv2023

Ensemble of deep learning language models to support the creation of living systematic reviews for the COVID-19 literature: a retrospective study

BackgroundThe COVID-19 pandemic has led to an unprecedented amount of scientific publications, growing at a pace never seen before. Multiple living systematic reviews have been developed to assist professionals with up-to-date and trustworthy health information, but it is increasingly challenging for systematic reviewers to keep up with the evidence in electronic databases. We aimed to investigate deep learning-based machine learning algorithms to classify COVID-19 related publications to help scale-up the epidemiological curation process. MethodsIn this retrospective study, five different pre-trained deep learning-based language models were fine-tuned on a dataset of 6,365 publications manually classified into two classes, three subclasses and 22 sub-subclasses relevant for epidemiological triage purposes. In a k-fold cross-validation setting, each standalone model was assessed on a classification task and compared against an ensemble, which takes the standalone model predictions as input and uses different strategies to infer the optimal article class. A ranking task was also considered, in which the model outputs a ranked list of sub-subclasses associated with the article. ResultsThe ensemble model significantly outperformed the standalone classifiers, achieving a F1-score of 89.2 at the class level of the classification task. The difference between the standalone and ensemble models increases at the sub-subclass level, where the ensemble reaches a micro F1-score of 70% against 67% for the best performing standalone model. For the ranking task, the ensemble obtained the highest recall@3, with a performance of 89%. Using an unanimity voting rule, the ensemble can provide predictions with higher confidence on a subset of the data, achieving detection of original papers with a F1-score up to 97% on a subset of 80% of the collection instead of 93% on the whole dataset. ConclusionThis study shows the potential of using deep learning language models to perform triage of COVID-19 references efficiently and support epidemiological curation and review. The ensemble consistently and significantly outperforms any standalone model. Fine-tuning the voting strategy thresholds is an interesting alternative to annotate a subset with higher predictive confidence.

bioinformatics↗

bioRxiv2022

Geographically Weighted Linear Combination Test for Gene Set Analysis of a Continuous Spatial Phenotype as applied to Intratumor Heterogeneity.

BackgroundThe impact of gene-sets on phenotype is not necessarily uniform across different locations of a cancer tissue. This study introduces a computational platform, GWLCT, for combining gene set analysis with spatial data modeling to provide a new statistical test for association of phenotypes and molecular pathways in spatial single-cell RNA-seq data collected from an input tumor sample. MethodsAt each location, the most significant linear combination is found using a geographically weighted shrunken covariance matrix and kernel function. Whether a fixed or adaptive bandwidth is determined based on a cross validation procedure. Our proposed method is compared to the global version of linear combination test (LCT), bulk and random-forest based gene-set enrichment analyses using data created by the Visium Spatial Gene Expression technique on an invasive breast cancer tissue sample, as well as 144 different simulation scenarios. ResultsIn an illustrative example, the new geographically weighted linear combination test, GWLCT, identifies the cancer hallmark gene-sets that are significantly associated at each location with the five spatially continuous phenotypic contexts in the tumors defined by different well-known markers of cancer-associated fibroblasts. Scan statistics revealed clustering in the number of significant gene-sets. A spatial heatmap of combined significance over all selected gene-sets is also produced. Extensive simulation studies demonstrate that our proposed approach outperforms other methods in the considered scenarios, especially when the spatial association increases. ConclusionsOur proposed approach considers the spatial covariance of gene expression to detect the most significant gene-sets affecting a continuous phenotype. It reveals spatially detailed information in tissue space and can thus play a key role in understanding contextual heterogeneity of cancer cells.

bioinformatics↗

bioRxiv2021

Information retrieval in an infodemic: the case of COVID-19 publications

The COVID-19 pandemic has led to an exponential surge and an enormous amount of published literature, both accurate and inaccurate, a term usually coined as an infodemic. In the context of searching for COVID-19 related scientific literature, we present an information retrieval methodology for effectively finding relevant publications for different information needs. Our multi-stage information retrieval architecture combines probabilistic weighting models and re-ranking algorithms based on neural masked language models. The methodology was evaluated in the context of the TREC-COVID challenge, achieving competitive results with the top ranking teams participating in the competition. Particularly, the ranking combination of bag-of-words and language models significantly outperformed a BM25-based baseline model (16 percentage points for the NDCG@20 metric), correctly retrieving more than 16 out of the top 20 documents retrieved. The proposed pipeline could thus support the effective search and discovery of relevant information in the case of an infodemic.

bioinformatics↗

bioRxiv2020

Identification of the Molecular Basis of Anti-fibrotic Effects of Soluble Guanylate Cyclase Activator Using the Human Lung Fibroblast Phosphoproteome

Idiopathic pulmonary fibrosis (IPF) is an irreversible and progressive fibrotic lung disease. Advanced IPF patients often demonstrate pulmonary hypertension, which severely impairs patients quality of life. The critical physiological roles of soluble guanylate cyclase (sGC)-cyclic guanosine monophosphate (cGMP) pathway have been well characterized in vasodilation and the corresponding therapies and pathway agonists have shown clinical benefits in treating hypertension. In recent years, many preclinical studies have demonstrated anti-fibrotic efficacy of sGC-cGMP activation in various experimental fibrosis models but the molecular basis of the efficacy in these models are not well understood. Also, sGC pathway agonism has demonstrated encouraging clinical benefits in advanced IPF patients (NCT00517933). Here, we have revealed the novel phosphorylation events downstream of sGC activation in human lung fibroblasts using phosphoproteomics. sGCact A, a potent and selective sGC activator, significantly attenuated more than 2,000 phosphorylation sites. About 20% of phosphorylation events, attenuated by transforming growth factor {beta} (TGF{beta}), a master regulator of fibrosis, were further dysregulated in the sGCact A co-treated lung fibroblasts. The overall magnitude and diversity of the sGCact A phosphoproteome was extensive. Further investigation would be required to understand how these newly identified changes facilitate human pulmonary fibrosis.

molecular biology↗

bioRxiv2019

Differential stromal reprogramming in benign and malignant naturally occurring canine mammary tumours identifies disease-promoting stromal components

The importance of cancer-associated stroma (CAS) for initiation and progression of cancer is well accepted. However, as stromal changes in benign forms of naturally occurring tumours are poorly understood, it remains unclear how CAS from benign and malignant tumours compare. Spontaneous canine mammary tumours are viewed as excellent models of human mammary carcinomas (mCA). We have recently reported highly conserved stromal reprogramming between canine and human mCA based on transcriptome analysis of laser-capture-microdissected FFPE specimen. To identify stromal changes between benign and malignant mammary tumours, we have analysed CAS and matched normal stroma from 13 canine mammary adenomas and compared them to 15 canine mCA. Our analyses revealed distinct stromal reprogramming even in small benign tumours. While similarities in stromal reprogramming exist, the CAS signature clearly distinguished adenomas from mCA, suggesting that it may reliably discriminate between benign and malignant tumours. We identified strongly discriminatory genes and found strong differential enrichment in several hallmark signalling pathways between benign and malignant CAS. The distinction between CAS from adenoma and mCA was further substantiated by differential abundance in cellular composition. Finally, to determine key players in CAS reprograming between adenomas and mCA, a network-based gene screening method identified modules of co-expressing genes with distinct expression profile in benign and malignant CAS, and revealed several hub genes as potential molecular drivers in CAS. Given the relevance of canine CAS as a model for the human disease, our approach identifies potential stromal drivers of tumour malignancy with implications for human mCA.\n\nSummary statementRNAsequencing-based analysis of stromal reprogramming between benign and malignant naturally occurring canine mammary tumours identifies potential molecular drivers in cancer-associated stroma that support tumour growth and malignancy.

cancer biology↗