# Why one DNA change can have several predicted consequences

Draft • Research checked 4 October 2026 • AI-assisted DomDNA editorial content. No independent clinical review or publication is claimed. General education, not individual medical advice.

An annotation table can give the same DNA change several labels: one row says intronic, another says missense, and a third names a different transcript. It can look as though the software cannot decide what happened. Often, the rows describe the change against different RNA transcripts from the same region.

A transcript is not another person's result. It is a particular RNA product or reference representation used to describe gene activity and sequence. The choice affects what a predicted consequence means.

## Follow one location through two transcripts

The [NHGRI RNA fact sheet](https://www.genome.gov/about-genomics/educational-resources/fact-sheets/ribonucleic-acid-fact-sheet) explains that alternative splicing can produce different RNA molecules from the same gene. Some sequence can be included in one transcript and excluded in another.

Consider a hypothetical annotation, with no real gene or patient implied. Transcript A includes a particular stretch in its protein-coding sequence. Transcript B excludes that stretch from its mature coding product. A change at that location could therefore receive a coding-consequence label for A and a different label for B.

There is no need to choose the most frightening row as the “true” answer. You need to know which transcript each row describes and why that transcript is relevant to the question.

The [Ensembl consequence documentation](https://mart.ensembl.org/info/genome/variation/prediction/predicted_data.html) explicitly describes predictions for allele-and-transcript combinations. Its consequence ordering is a tool convention. The label is not a measurement of symptoms or a diagnosis in the person whose data is annotated.

## What a transcript identifier buys you

An identifier lets another reader recover the reference representation used for the statement. Keep the version where available. If you save only the gene name and a coloured impact label, you remove much of the information needed to compare annotations.

A helpful note says: “This consequence was predicted for this transcript in this annotation release.” It does not say: “This gene is damaged.” The second sentence adds biological and clinical conclusions that the annotation may not establish.

The same care applies to “most severe consequence” summaries. They can be useful for reducing a long table, but a summary cannot preserve every transcript-specific distinction. A display choice is not a finding that the selected transcript is always the important one in every tissue or condition.

## A common reference is helpful, not all-encompassing

The [NCBI and EMBL-EBI MANE project](https://www.ncbi.nlm.nih.gov/refseq/MANE/) provides matched transcript references. MANE Select offers a representative transcript, while MANE Plus Clinical includes additional transcripts where needed for relevant clinical reporting.

This is a practical effort to make communication more consistent. It does not imply that a gene has only one transcript, that other RNA products are imaginary or that one annotation field settles pathogenicity.

When two reports use different transcript references, disagreement in a consequence label may be an annotation issue worth resolving before discussing clinical meaning. The underlying DNA observation need not have changed.

## Read a table without overreading it

Start with the column headings. Separate the observed sequence change from the predicted consequence, the transcript reference and any clinical classification. If those appear in a single coloured badge, look for the detailed view or report notes.

In a hypothetical conversation, “Why does this row say intronic while that row says missense?” is a precise question. “Which transcript does each row use, and which one is relevant here?” moves it towards a useful answer. “Which colour should I worry about?” leaves the important distinction hidden.

Your practical next step is to keep the transcript identifier alongside one annotation you want explained. Preserve the original table rather than cropping out all but the impact label. Do not use a computational consequence alone to decide on treatment or tell relatives they have a condition.

Once the reference is visible, multiple rows become easier to understand. They may be descriptions of several possible RNA contexts, rather than several conflicting diagnoses.

Here is a hypothetical comparison task: one export retains both an identifier and its version, while another retains only the identifier. They have not supplied equally detailed annotation context. Mark the second version as unknown rather than assuming it matches. That keeps a technical omission visible without inventing a clinical disagreement between the files.

For general lifestyle learning, [DomDNA’s educational quiz](https://domdna.com/quiz) does not annotate transcripts or analyse genetic files.