Molecular Sciences - Posters L
Szilvia Barsi, Bence Szalai, László Hunyady
1 Semmelweis University, Faculty of Medicine, Department of Physiology, Budapest, Hungary
2 Semmelweis University, Faculty of Medicine, Department of Physiology, Budapest, Hungary,
Research Centre for Natural Sciences, Institute of Enzymology, Budapest, Hungary
3 Semmelweis University, Faculty of Medicine, Department of Physiology, Budapest, Hungary, MTA-SE Laboratory of Molecular Physiology, Budapest, Hungary Research Centre for Natural Sciences, Institute of Enzymology, Budapest, Hungary
Intercellular communication is a fundamental process, where the activated receptors initiate the downstream signalling, which is essential for the adaptation to their environment. The detection of activated receptors is crucial for understanding the molecular mechanisms that regulate cellular processes. There are computational methods for inferring receptor or ligand signalling activities. Data-driven approaches tend to outperform the prior-knowledge-based methods however they can be more interpretable. Combining receptor and ligand perturbation gene expression signatures and the known ligand-receptor interactions overcome this limitation.
The aim of this study was to develop a large-scale statistical model to reliably estimate receptor activities from gene expression profiles of independent samples.
We collected the receptor and ligand perturbation gene expression signatures from the LINCS L1000 database and used ligand-receptor interactions described in the literature to construct a database containing 38989 unique transcriptional signatures for 599 receptors. We developed a statistical model that describes the relationship between the receptors and the altered gene expression patterns using a linear regression model. We investigated the correlation between predicted receptor activities and baseline expression of receptors and ligands of different cell lines from The Cancer Genome Atlas (TCGA) and Cancer Cell Line Encyclopedia (CCLE). We benchmarked the performance of our model by comparing how well they can recover the perturbed receptors using receiver operating characteristic analysis.
We have developed a model that can infer receptor activities from transcriptional profiles independent of sample type and conditions. The correlation analysis in TCGA and CCLE revealed that the constructed database reflects downstream signalling of receptors in independent human tissues. Comparing the performance with state-of-art methods, like CytoSig, that infers cytokine signalling activities, our model exhibited an improvement in recovering which cytokine was perturbed in the sample.
We have demonstrated that our large-scale model infers receptor activities reliably from gene expression thus it helps to understand the communication-related mechanisms behind transcriptional patterns.
ÚNKP-22, Nemzeti Laboratóriumok Program