RNA sequencing can examine many transcripts in one experiment. The resulting table may contain thousands of estimates, yet its interpretation still begins with a specific comparison between samples. Read the sample design and processing before treating a list of changed genes as an explanation of what a peptide does.
Identify the comparison the analysis actually tests
Write down the cell or tissue model, peptide condition, reference condition and sampling time. If the study includes several doses or intervals, identify which contrast produced the table you are reading.
For an original fictional example, a dataset contains untreated samples at two times and exposed samples at the later time. A comparison between exposed and untreated late samples differs from a comparison between exposed late samples and untreated early samples.
The latter changes time as well as exposure. A result from that contrast should not silently become an exposure-only claim unless the analysis and design support the separation.
ENCODE’s bulk RNA-seq standards distinguish biological replication and library characteristics as part of experiment assessment. These details help identify what material and variation the dataset represents; the standard is not a guarantee about every independently published dataset.ENCODE — Bulk RNA-seq Data Standards (opens in a new tab)
Separate sequencing depth from biological abundance
Love and colleagues’ DESeq2 methods paper describes modelling RNA-seq counts with normalisation factors and estimates of variability between samples. Raw counts are therefore inputs to the analysis, rather than final biological comparisons requiring no adjustment.Love, Huber and Anders — Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2 (opens in a new tab)
| Library | Reads assigned to feature | Total reads |
|---|---|---|
| A | 100 | 1 million |
| B | 200 | 2 million |
The feature has the same simple fraction of total reads in both fictional libraries. Its doubled raw count alone does not establish doubled biological abundance. This table is an illustration of sampling scale, not a recommendation to normalise real RNA-seq data by this ratio.
Read the method the authors actually used, including its assumptions. Different normalisation approaches and sample compositions can affect the comparison. A software name should lead to a method description, not replace one.
Read effect size and uncertainty together
The DESeq2 paper explains how information shared across genes can stabilise dispersion and fold-change estimates, particularly where data are limited. This is a modelling choice with a rationale; it does not make every numerical estimate equally informative.Love, Huber and Anders — Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2 (opens in a new tab)
Find the reported effect size, uncertainty or statistical evidence, and any expression-level or filtering information. A gene list sorted by a significance measure answers a different ordering question from one sorted by estimated fold difference.
For a fictional result, a large estimated change with weak supporting information should retain that uncertainty. Conversely, a precisely estimated small change should not be described as large merely because its statistical evidence is strong.
Check whether the table addresses multiple comparisons and whether missing entries were filtered, unmeasured or not statistically selected. Absence from a highlighted list does not establish absence of expression or an exactly unchanged transcript.
Keep the transcript list separate from functional activation
A collection of changed transcripts can suggest biological themes. If the paper reports pathway enrichment, identify the analysis and its background set rather than treating the pathway label as a direct activity measurement.
The list still concerns RNA measured in the sampled material. It does not automatically specify protein abundance, phosphorylation, metabolic flux or the direction of a whole pathway’s functional response.
Read follow-up measurements for those additional claims. A targeted transcript check can corroborate selected RNA findings, while a protein or functional experiment addresses a different level of evidence.
A clear summary names the contrast, model, time point and principal transcript finding, then describes downstream interpretations as tested outcomes or hypotheses as appropriate. That gives readers a usable evidence map without turning a long gene list into a complete mechanism.
Sources and further detail
- Love, Huber and Anders — Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2 (opens in a new tab)
Primary 2014 methods paper read for count modelling, normalisation and information sharing. No claim that historical software defaults equal current defaults; no published dataset results reproduced.
- ENCODE — Bulk RNA-seq Data Standards (opens in a new tab)
Indexed consortium standards consulted for replication and library-characteristic reporting. Standards describe ENCODE assessment and do not independently validate a peptide dataset.
Sources checked 19 September 2026. Worked examples are illustrative unless a supplied report is explicitly identified. This article has not undergone independent scientific peer review.