Today we have powerful tools to engineer biological systems by changing the
DNA 'source code'. However, biological systems are extraordinarily complex
and knowing which changes to make to bring about a target result is an
ongoing problem. With increasing amounts of high-throughput 'big data'
being produced, Deep Learning is becoming a promising technology to improve
our predictive performance and reduce the effort intensive wet-lab work
required to test biological hypotheses. Here, we specifically focus on
one sub-problem, structural gene prediction. Our prototype Deep Learning
models make ground breaking progress over the existing Hidden Markov Model
based tools (https://doi.org/10.1093/bioinformatics/btaa1044).
Ongoing work focuses on further improving performance, capturing the
full biological complexity and developing applicability.