Login / Signup

Transcriptome Complexity Disentangled: A Regulatory Molecules Approach.

Amir AsiaeeZachary B AbramsKevin R Coombes
Published in: bioRxiv : the preprint server for biology (2023)
Gene regulatory networks play a critical role in understanding cell states, gene expression, and biological processes. Here, we investigated the utility of transcription factors (TFs) and microRNAs (miRNAs) in creating a low-dimensional representation of cell states and predicting gene expression across 31 cancer types. We identified 28 clusters of miRNAs and 28 clusters of TFs, demonstrating that they can differentiate tissue of origin. Using a simple SVM classifier, we achieved an average accuracy of 92.8% in tissue classification. We also predicted the entire transcriptome using Tissue-Agnostic and Tissue-Aware models, with average R 2 values of 0.45 and 0.70, respectively. Our Tissue-Aware model, using 56 selected features, showed comparable predictive power to the widely-used L1000 genes. However, the model’s transportability was impacted by covariate shift, particularly inconsistent microRNA expression across datasets.
Keyphrases
  • gene expression
  • single cell
  • rna seq
  • transcription factor
  • dna methylation
  • genome wide
  • poor prognosis
  • cell therapy
  • deep learning
  • high speed
  • high resolution
  • mesenchymal stem cells
  • long non coding rna