Introduction
Enhancers are non-coding DNA sequences that control gene expression and are essential for defining cell identity and function. They are a major focus in pharma and biotech because understanding and targeting enhancer activity can improve gene therapy design, cell reprogramming, and the interpretation of disease-associated variants. Predicting where enhancers are located and which genes they regulate has become a central challenge in genomics. The field has produced many computational models that attempt to predict enhancer activity from DNA sequence, chromatin accessibility, and other epigenetic features.
Until recently, there was no clear way to evaluate how well these models actually perform. Most were tested on different datasets with limited validation of whether their predicted enhancers were truly functional in biological systems.
A new study in Cell Genomics provides an important advance by introducing a benchmark dataset of validated enhancers and using it to evaluate a wide range of prediction tools. This work moves enhancer prediction from a collection of competing models to a more unified and testable field.
What This Study Adds
To test how well current tools predict enhancers, the authors created a high-quality set of known enhancers from the mouse brain and compared two main types of models.
Chromatin-based models use data like ATAC-seq to find regions of DNA that are open and accessible, which often marks active enhancers. Sequence-based models use machine learning to scan DNA for patterns that suggest enhancer function, without needing lab data.
The authors found that chromatin-based models were the most accurate way to find active enhancers. But the sequence-based models were still valuable. They helped reduce false positives (regions that look like enhancers but don’t actually function as such), and offered insight into the regulatory logic that defines how enhancers work in different cell types. This complementary role is especially useful when chromatin data alone overestimates enhancer activity. Sequence models help refine predictions by asking not just is the region open but also does the sequence look like an enhancer.
Going forward, combining these approaches could improve how we identify functional enhancers, interpret non-coding regions of the genome, and design synthetic enhancers for research or therapeutic use.
Why This Matters
This study gives the field its first rigorous, standardized way to compare enhancer prediction tools. It helps researchers understand the strengths and limitations of each approach. Chromatin-based models perform well at identifying regions with regulatory potential. Sequence-based models help distinguish functional from non-functional sites and reveal design rules that can guide synthetic biology.
By using this benchmark, developers of new prediction methods can test their models against a common reference and build tools that are easier to compare and interpret. This helps move the field toward more reliable and reproducible enhancer prediction.
Looking Ahead
Enhancer prediction is becoming a practical tool for genome interpretation, gene therapy, and synthetic regulatory design. With a standardized benchmark now available, the field is better positioned to improve these models, evaluate them fairly, and apply them more confidently across research and clinical settings.
Bridge Informatics supports this kind of work by helping clients apply enhancer prediction models to real-world datasets through expert data integration and bioinformatics analysis. Click here to schedule a free introductory call with a member of our team.
