Enformer
A deep learning model for genomic sequences developed by Google DeepMind. It is an important predecessor to AlphaGenome and subsequent work, predicting functional characteristics such as gene expression from genomic sequence inputs.
Timeline (last 90 days)
Similar models such as DeepSEA, Enformer, and Sei perform well on known variants, but their generalizability to novel variants outside the training distribution or to extremely rare ethnic groups is questioned.
The AlphaGenome model drew inspiration from earlier work such as Enformer.
Enformer, released by Google DeepMind in 2021, uses a Transformer architecture with a context window of approximately 200 kb.
DeepMind's previous Enformer model has an effective input window of approximately 200 kb.
All Facts (4)
Similar models such as DeepSEA, Enformer, and Sei perform well on known variants, but their generalizability to novel variants outside the training distribution or to extremely rare ethnic groups is questioned.
50%UnverifiedThe AlphaGenome model drew inspiration from earlier work such as Enformer.
50%UnverifiedEnformer, released by Google DeepMind in 2021, uses a Transformer architecture with a context window of approximately 200 kb.
50%UnverifiedDeepMind's previous Enformer model has an effective input window of approximately 200 kb.
50%