Authors: Eric Zhuang, Manoj Kandpal
Ribosome profiling sequences messenger RNA (mRNA) fragments protected by ribosomes. Each sequenced fragment is represented as a read. Where a read aligns shows the region occupied by a ribosome, but it does not directly show which nucleotide lies at the ribosome's peptidyl site (P site). Identifying that nucleotide is necessary for analyses of individual codons, the three-nucleotide units read during translation. Existing P-site predictors require different inputs, define positions differently, report either one position per read or a signal summarized across many reads, and vary in whether they use information from the experimental sample being evaluated. These differences make fair comparison difficult.We developed TabRiboICL, a novel predictor that assigns one P-site position to each read. It evaluates 13 possible positions located 8—20 nucleotides from the read's 5u2032 end. The published TabICL machine-learning model [10] scores each possible position. TabRiboICL combines each score with how often the corresponding distance occurred among similar training reads with expected P-site positions and then selects one P-site position. We also developed a benchmarking framework that converts the different forms of output to the same definition of a read-level P-site position and evaluates every predictor on the same reads.The benchmark includes 8 public datasets, 31 ribosome-profiling samples from Drosophila melanogaster and Saccharomyces cerevisiae, and 11 predictors. An expected P-site position was assigned to a read near the annotated beginning or end of a protein-coding region only when exactly one of the 13 possible positions matched the relevant boundary and fell on the first nucleotide of an annotated codon. Because these expected positions were inferred rather than directly observed, we call them proxy labels. Exact accuracy, the primary measure, required a prediction to match its proxy label. Accuracy within one nucleotide was the secondary measure.During the held-out dataset experiment, TabRiboICL had the highest observed mean exact accuracy after every dataset was given equal weight: 37.6% (95% bootstrap confidence interval, 32.5%—43.8%). The interval was calculated by repeatedly redrawing datasets, samples, and reads. TabRiboICL had the highest observed exact accuracy in 5 of 8 datasets; Scikit-ribo had the highest in two datasets and LightGBM in one. The framework also characterized predictor performance across datasets, organisms, and recorded protocol categories, providing a consistent, well-documented basis for comparing P-site predictors.
Comments: 32 Pages.
Download: PDF
[v1] 2026-11-15 01:47:35
Unique-IP document downloads: 0 times
Vixra.org is a pre-print repository rather than a journal. Articles hosted may not yet have been verified by peer-review and should be treated as preliminary. In particular, anything that appears to include financial or legal advice or proposed medical treatments should be treated with due caution. Vixra.org will not be responsible for any consequences of actions that result from any form of use of any documents on this website.
Add your own feedback and questions here:
You are equally welcome to be positive or negative about any paper but please be polite. If you are being critical you must mention at least one specific error, otherwise your comment will be deleted as unhelpful.