FAQs

Frequently asked questions

Documentation · Tutorial

How do I choose between KEA and KRS?

Use KEA to compare two RNA populations. Use KRS to characterize one RNA relative to a population. Either signature can be passed to K-RBP.

Is K-tools a web server?

K-tools is a Python library.

Does it accept RNA sequences containing U?

Convert U to T and uppercase before FASTA counting. The enumerated count universe is A/C/G/T. K-RBP can convert T to U to match RNA-formatted score matrices.

Does it count overlapping occurrences?

Yes, but the reference-class counter omits the final possible window: AAAA, k=2 yields two AA counts. Use sequences longer than k.

Is a k=7 analysis always appropriate?

No. Choose k according to sequence length, data size and sparsity. A 4ᵏ universe grows exponentially. The bundled tutorial uses k=3 for a small, fast demonstration. Sparse k=7 populations can have zero quartile spread and undefined delta-median scores.

Is sign_emp_pval a P-value cutoff?

For delta selections it is a tail fraction of sorted scores. Statistical significance is controlled separately by fdr_alpha and correction_method in stat_log2fc.

Why is my integrated signature empty?

The group-level and transcript-level selections may not overlap. This is a valid outcome. Inspect the full tables, group design, sparsity and thresholds. Do not drop the empty method and report the other method's list as the intersection.

Why is a k-mer missing from the statistical table?

The code skips k-mers for which both groups have constant normalized frequencies, even if the two constant values differ. This is an implementation limitation. If every k-mer is skipped, the current method can fail rather than return an empty table.

Does KRS top_pct=5 select 5% of k-mer types?

No. It selects k-mers for which the target ranks at or above the 95th percentile among transcripts containing that k-mer. Signature size varies. The target is included in the ranking.

Is zscore_diff a Z score?

In K-RBP it stores Cliff's delta. Positive values indicate higher query scores. The effect_size field has the opposite sign under the MWU branch; use zscore_diff for the documented biological filters.

Why is my K-RBP plot empty?

Check that profiles pass both the significance and effect-size filters. Set fc_thresh explicitly; the default 1.0 can exclude all points under a strict filter because Cliff's delta cannot exceed 1. A word cloud raises an error when no profiles pass. The full returned table is separate from plot selection.

Is the RBP score matrix included?

Two PEKA matrices are included in data/peka. Choose one input file. The synthetic demo uses a fictional matrix.

Are results deterministic?

Set NumPy's global seed before KEA control resampling. K-RBP uses seed 42 for background subsampling but set-derived ordering can change the sampled background between processes. Pin input ordering and package versions, and record the code commit.

Do the advanced mapping helpers run on my computer?

BEDTools helpers require bedtools on PATH. Genome mapping also requires the external mapper script and resources configured with KTOOLS_MAPPER_SCRIPT and KTOOLS_MAPPER_DIR.

What should I cite before the preprint is published?

Cite the repository URL and the exact release or commit used, with access date. The preprint DOI will be added when available.

Is uploading the wiki/ folder enough to enable GitHub Wiki?

No. GitHub Wiki is a separate Git repository. The folder contains prepared page sources; publish them through the Wiki editor or export them to a clone of the Wiki repository. The main docs/ pages work immediately after uploading the repository.