New statistical method improves detection of gene regulation and interaction networks

Understanding gene regulation—how cells control which genes turn on and off—remains one of genetics' central challenges. A new method tackles two interconnected problems at the same time: finding which genetic variants affect gene activity, and mapping the hidden network of interactions between genes themselves.

The challenge: two linked problems#

Scientists need to solve two major tasks in regulatory genomics. The first involves eQTL mapping, which identifies associations between specific DNA variants (SNPs) and changes in gene expression levels. The second is gene network estimation, which reveals how genes directly influence each other's activity. Traditionally, these have been treated as separate problems, missing an opportunity to leverage shared patterns between them.

A unified approach#

Researchers have developed a computational technique called $L_{1-2}$ GLasso—a regularized multi-task graphical lasso method. Rather than analyzing eQTL mapping and gene networks in isolation, this approach jointly estimates both simultaneously, allowing information to flow between the two tasks. This shared learning framework makes the method more efficient at identifying the true sparse patterns underlying gene regulation.

Validation and real-world results#

Testing on artificial data demonstrated the method's ability to accurately recover the expected sparse structures in both eQTL mapping and gene networks. When applied to real genomic data from the ADNI-1 study, $L_{1-2}$ GLasso produced sparser and more reliable solutions compared to other standard methods in the field. This suggests the approach could improve how researchers pinpoint functional genetic variants and regulatory relationships in disease-relevant tissues.