Terminitor: Cleavage Site Prediction Using Deep Learning Models
Bibliographic record
Abstract
Abstract As a widespread RNA processing machinery, alternative polyadenylation plays a crucial role in gene regulation. To help decipher its underlying mechanism and understand its impact, it is desirable to comprehensively profile 3’-untranslated region cleavage and associated polyadenylation sites. State-of-the-art polyadenylation site detection tools are known to be influenced by library preparation artefacts or manually selected features. Moreover, recently published machine learning methods have only been tested on pre-constructed datasets, thus lacking validation on experimental data. Here we present Terminitor, the first deep neural network-based profiling pipeline to make predictions from RNA-seq data. We show how Terminitor outperforms competing tools in sensitivity and precision on experimental transcriptome sequencing data, and demonstrate its use with data from short- and long-read sequencing technologies. For species without a good reference transcriptome annotation, Terminitor is still able to pass on the information learnt from a related species and make reasonable predictions. We used Terminitor to showcase how single nucleotide variations can create or destroy polyadenylated cleavage sites in human RNA-seq samples. Author Summary 3’ cleavage and polyadenylation of pre-mRNA is part of RNA maturation process. One gene can be cleaved at different positions at its 3’ end, namely alternatively polyadenylation, thus identifying the correct polyadenylated cleavage site (poly(A) CS) is essential to unveil its role in gene regulation under different physiological and pathological conditions. The current poly(A) CS prediction tools are either heavily influenced by RNA-Seq library preparation artefacts or have only been designed and tested on ad hoc datasets, lacking association with real world applications. In this study, we present a deep learning model, Terminitor, that predicts the probability of a nucleotide sequence containing a poly(A) CS, and validated its performance on human and mouse data. Along with the model, we propose a poly(A) CS profiling pipeline for RNA-seq data. We benchmarked our pipeline against competing tools and achieved higher sensitivity and precision in experimental data. The usage of Terminitor is not limited to genome and transcriptome annotation and we expect it to facilitate the identification of novel isoforms, improve the accuracy of transcript quantification and differential expression analysis, and contribute to the repertoire of reference transcriptome annotation.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".