AI model decodes DNA signature of transcription initiator in 60% of human genes
A machine-learning analysis of around 500,000 DNA sequences has characterised the initiator element — a core promoter signal — at unprecedented scale, offering a route to predicting the impact of non-coding mutations.
Researchers have used a machine-learning model to identify and characterise the DNA sequence signature of the initiator (Inr) element, a core component of eukaryotic promoters that is required to position RNA polymerase II at transcription start sites. Training on approximately 500,000 human DNA sequences, the model detected an initiator motif at roughly 60 per cent of human gene promoters — a substantially larger fraction than previous biochemical surveys had established.
The Inr element had been known since the early 1990s, but its precise sequence rules and genome-wide prevalence remained poorly defined, limiting the ability to predict whether variants in promoter regions affect gene expression. The new AI-assisted approach provides a more complete picture of where and how this switch operates across the genome.
The finding has potential relevance for interpreting the functional consequences of non-coding variants identified in population sequencing studies and disease genetics research. The authors suggest the model could eventually contribute to broader efforts to decode the cis-regulatory grammar that governs gene activity — a longstanding goal in functional genomics. Full institutional attribution, journal details, and authorship are not available from the ScienceDaily report; readers should consult the primary publication for methodological detail.
Note: this summary is based on a ScienceDaily press-release summary; the underlying journal article should be consulted before drawing research conclusions.
Plain-language version
For patients, families, and general readers. Educational only — not medical advice.
Every cell in your body contains the same DNA, but different genes are switched on in different cells at different times. Special short sequences in the DNA, called promoter elements, act like 'on switches' that tell a cell when to start reading a gene. One of these switches is called the initiator element.
Researchers have used an artificial intelligence tool to scan about half a million human DNA sequences and work out what this particular switch looks like — and where it sits — across the human genome. They found it near the start of roughly six in ten human genes.
This kind of work helps scientists understand how mutations in the parts of DNA that are not genes themselves can still affect how genes behave. That understanding may, in future, help researchers make better sense of genetic test results.
This is an educational summary, not medical advice. If anything here raises questions for you, please speak with your GP or a clinical professional.
Sources
Read the original reporting — these are the public sources this summary draws from.
-
Primary source ScienceDaily · 2026-08-23A hidden "on switch" in human DNA has finally been decoded