1. How to cite OneStopRNAseq?

Cite the OneStopRNAseq paper, then describe the analyses you actually ran. Start from the Template of Method Section and edit that text for your study. Do not paste the template unchanged.

Li, R.; Hu, K.; Liu, H.; Green, M.R.; Zhu, L.J. OneStopRNAseq: A Web Application for Comprehensive and Efficient Analyses of RNA-Seq Data. Genes 2020, 11, 1165. https://doi.org/10.3390/genes11101165

The template already contains suggested wording, software versions, and citations. Make two kinds of edits:

Insert the genome and annotation you actually selected. The template uses XXX as a placeholder, for example “Reads were aligned to the reference genome assembly XXX … and annotated with XXX.” Replace those placeholders with the assembly and annotation from your run. The genomes and annotations currently available are listed in the Template of Method Section, which has the latest information.

Also replace the other XXX placeholders with the values you entered: featureCounts parameters, the FDR cutoff (MAX_FDR), and the minimum absolute log2 fold change (MIN_LFC).

Keep only the analyses you actually performed with OneStopRNAseq. Delete every sentence for a step you did not run, and keep the citations only for the tools that remain. Typical blocks in the template are quality control (FastQC, MultiQC, and QoRTs), alignment (STAR), gene counting (featureCounts), differential expression (DESeq2), gene set enrichment (GSEA), alternative splicing (rMATS), differential exon usage (DEXSeq), transposable-element quantification (SalmonTE), and allele-specific expression (ASEReadCounter). For example, if you did not run alternative splicing, omit the rMATS sentence; if you started from a count table, omit alignment and counting; if you did not request transposable-element or allele-specific analysis, omit those paragraphs.

2. Why do some pathways have very similar GSEA significance (NES/FDR) across datasets, or a higher NES in the dataset with smaller gene-level fold changes, while individual genes in those pathways show much larger expression changes in one dataset than another?

GSEA and individual gene expression statistics measure different biological properties. GSEA evaluates whether genes within a pathway are collectively enriched near the top or bottom of a ranked gene list, whereas gene-level analyses focus on the magnitude of expression change for individual genes.

NES is driven primarily by the ranks of genes in the ordered list rather than the absolute magnitude of their fold changes. A pathway can achieve a high NES even when the individual fold changes are modest, provided that many pathway genes are consistently ranked near the top of the list. A dataset in which many pathway genes are consistently enriched near the top of the ranking can therefore produce a higher NES than a dataset containing a few genes with very large fold changes, if the remaining pathway genes are scattered throughout the ranked list.

In other words, NES reflects coordinated pathway-wide behavior, while log2 fold change reflects the magnitude of change for individual genes. NES and individual gene log2 fold change are therefore not expected to be tightly coupled and can diverge substantially.

3. What is the correct interpretation when NES and leading-edge gene expression appear inconsistent?

The apparent discrepancy is not a contradiction. NES quantifies the strength of enrichment of an entire pathway, whereas leading-edge genes highlight the subset of genes contributing most to that enrichment. These metrics describe related but distinct aspects of the data and therefore may not increase or decrease together.