Computational dissection of the sequence elements defining human cleavage and polyadenylation sites and their role in determining gene ends

After decades of research, significant progress towards understanding how living organisms decode their genomes has been made. Eukaryotes utilize protein factors as well as DNA sequence elements to define eukaryotic gene starts and gene ends. During 3′-end gene definition, when pre-mRNA is being produced, it has to go through several steps of maturation, including cleavage and polyadenylation (CPA), when the pre-mRNA gets cleaved and a string of ~200 adenine nucleotides gets added at 3′ end of the transcript. Mechanistically, CPA is thought to utilize five key sequence recognition elements that reside in pre-mRNA 3′-UTRs: the upstream UGUA, polyadenylation signal (PAS), U-rich sequences, the CA/UA dinucleotide where cleavage occurs, and GU-rich downstream elements (DSE). It remains contentious whether these sequences are sufficient to delineate CPA sites, however. To date, no studies have explicitly attempted to assess how well the established features differentiate CPA sites from the remainder of the genes. Furthermore, it is not clear what role these sequences play in defining human geneends. Here, I expand on previous work by dissecting the contributions of numerous individual sequence features to CPA site selection. Using classical discriminative approaches, I show that computational models can be constructed that accurately identify constitutive CPA sites, relative to the entire pre-mRNA sequence, for 41% of all genes. U1-hybridizing sequences, which have been described as one of the strong additional cryptic CPA regulators, have surprisingly little impact on model performance, suggesting their limited role in regulating human CPA. Addition of all known RBP motifs to the model, however, increases this figure to 49%, with feature scores that highlight known and suspected CPA regulators, as well as many potential new factors. Together, these results suggest that while the established features have a higher specificity in defining human CPA sites than expected, the existing knowledge of sequence elements contributing to CPA definition does not account for the full complexity of CPA regulation.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.