A research team from Cold Spring Harbor Laboratory has introduced new computational methods to predict transcription elongation rates using nascent RNA sequencing data. Transcription elongation is the process where RNA polymerase moves along DNA, synthesizing RNA. The speed of this process varies depending on different factors within the gene and surrounding regulatory elements. Understanding these variations can provide insights into gene regulation and potential impacts on disease.

The researchers developed a probabilistic model that predicts how fast transcription occurs at specific nucleotide positions by analyzing genomic and epigenomic features. They validated their model using simulations and applied it to existing PRO-seq (Precision Run-On Sequencing) and epigenomic datasets from four different cell types. Their findings showed that transcription slows down at specific sites, such as cytosine-rich regions, DNA methylation sites, splice sites, RNA stem-loops, and binding sites for the CTCF protein. In contrast, transcription speeds up in regions with thymine nucleotides, A+T-rich sequences, and specific histone modifications like H3K79me2.

To further improve their predictions, the team introduced a convolutional neural network, a type of machine-learning model that enhances accuracy by recognizing complex patterns in DNA sequences. This study is the first to provide genome-wide predictions of nucleotide-specific transcription elongation rates, offering valuable tools for studying gene regulation and potential applications in disease research.

Availability – The predictions for elongation rate are available in a UCSC Genome Browser track at http://compgen.cshl.edu/elongation-rate-tracks.php.