Enformer

Part of speech: noun

Definitions

1. A deep neural network that predicts gene activity and regulatory effects directly from raw DNA sequence data
2. A machine learning model trained to estimate how strongly genes are expressed based on the DNA sequence surrounding them
3. An artificial intelligence system used in genomics to forecast gene expression levels from nucleotide sequences alone

Etymology: "Enformer" is a blend of "enhancer" and "transformer." An enhancer is a stretch of DNA that boosts a gene's activity, often from many thousands of bases away along the chromosome. A transformer is the attention-based neural network design behind many modern language and image systems. The name points to the model's central problem: connecting distant regulatory DNA to the genes it controls. The blend is a deliberate coinage by the researchers who built the model, in the style of many technical names that fuse two terms to show a tool's ingredients at a glance. The system takes only a DNA sequence as input and predicts expression levels for thousands of genes, which is what distinguishes it from earlier methods that relied on short windows of sequence around each gene.