the issue you posted is related to

codon_coverage

not

cds_coverage

function

so we're fine, there's no issue here

riboWaltz works as expected, and the docs are fine, no need to chase

of course, if you wish, you can go into the functions and check if they work as expected, but in my experience with riboWaltz, I think the

cds_coverage

function does indeed only count in-frame P-sites

This will make the counts useful within sample but not between samples because the proportion of reads we expect to see in frame 0 of the CDS will change with this property

@Jack Tierney Would you argue that this is fundamentally different than any other bias that is inherent to a sample, thus making inter-sample comparisons pointless? The same argument could be made about RNA-seq samples having different biases. No RNA-seq sample will ever be like the next one, even if they are prepared in the same batch. Yet, it is common practice to group them together and perform differential expression analysis. Would you say this is not valid for Ribo-seq data?

One idea behind counting in-frame p-sites is to identify overlapping ORFs shifted by only one or two bases. This may be a microprotein-specific question, which is where the idea comes from, and not applicable to a wider Ribo-seq audience.

From the feedback I have received from colleagues, I have understood that this method of counting is mostly relevant to the study of non-canonical ORFs coding for microproteins. They agree that the issues that you brought up in this Slack conversation complicate the interpretation, but they say the quantification is useful in the context of microproteins nonetheless. I have amended the documentation to reflect this specialized use case and also pointed out the caveats associated with the method. The warnings should safeguard against misuse of the method. With these warnings in place, would you feel more comfortable incorporating the feature into the pipeline?

Hi Jack, Regarding the frame restricted quantification: In the context of the non-canonical translatome, frame is far more than a QC metric—it is the defining feature of the biological signal. While standard total-count methods are robust for well-separated canonical CDSs, they collapse the frame dimension, making them fundamentally incapable of distinguishing the overlapping ORFs that characterize the hidden landscape of the dark proteome. The primary justification for this method is its unique ability to distinguish between multiple translation events occurring on the same transcript. In the non-canonical translatome, we frequently encounter: • Overlapping ORFs: Dual-coding regions where one sequence translates two different proteins in different frames (e.g., a canonical protein in frame 0 and a microprotein in frame +1). • Upstream and Downstream Overlapping ORFs (uoORFs and doORFs): Which physically overlap with the boundaries of the main CDS. • Internal ORFs: Microproteins encoded entirely within the sequence of a larger CDS but in a different reading frame. Without frame-specific P-site counting, the signal from a highly translated canonical protein would mask the signal of a lowly translated overlapping microprotein. By restricting quantification to the fixed in-frame position (Frame 0) relative to each specific ORF, we apply a biological filter that isolates the relevant translation from the surrounding 'noise' of other reading frames This approach is not a fringe methodology; it is a specialized tool used by researchers specifically to gain the resolution needed for microprotein discovery and quantification. Landmark papers identifying thousands of novel ORFs—such as van Heesch et al. (2019, Cell), Hofman et al. (2024, Mol Cell) , and Ruiz-Orera, J., et al. (2024, Nat Cardiovasc Res)—rely on frame-restricted quantification.

Technical Note: Assumptions and Limitations of Frame-Specific Counting While frame-0 P-site quantification is essential for resolving overlapping or non-canonical ORFs, users must account for "frame-0 capture efficiency" when performing inter-sample comparisons. Unlike standard RNA-seq counts, which are largely independent of transcript structure, frame-specific counts are sensitive to the global periodicity of the library. If an experimental condition or library preparation method alters the distribution of ribosome footprints (e.g., changing the ratio of in-frame to out-of-frame reads), the resulting changes in counts may reflect shifts in P-site assignment accuracy rather than true changes in translational output. This is particularly relevant in treatments that perturb ribosome movement, such as elongation inhibitors or cellular stress, which can "blur" the 3-nucleotide periodicity and lead to signal leakage between frames. To mitigate these risks, it is strongly recommended to use this quantification method in conjunction with rigorous periodicity QC metrics (such as those provided by

riboWaltz

or

multiqc

). If the percentage of in-frame P-sites varies significantly between experimental groups, differential expression results should be interpreted as a composite of translational regulation and frame-occupancy shifts. For studies focused on microproteins or overlapping ORFs, this method remains the most viable path to signal isolation; however, users should validate that their findings are not artifacts of capture efficiency by verifying that global frame distribution remains stable across the compared biological replicates.

Let me know what you think please. Thank you!

hi, I am on annual leave and wont be able to review the changes to this PR. i'd highly encourage a review from a maintainer, maybe Jon is available now?

Reviewed. Main thing is that I think the looping within the new process should be delegated to nextflow for scalability.