I tried to write a deep neural network model into a qza file and I was wondering if there’s any built-in function that might be useful.
Thanks.
Maybe, @gcj13 but that depends on what you mean by built-in functions!
The QIIME 2 framework doesn’t know anything about informatics. This is an intentional decision, and means that your plugin can use whatever tools it needs, and define whatever data types and formats it needs to do its work.
As such, QIIME 2 doesn’t offer built-in functions for neural networks. It does, however, offer APIs for defining, validating, and transforming semantic types and formats, which are important to how the framework and other plugins handle .qzas. There are also good real-world examples of these definitions in q2-types, q2-feature-classifier, and many other plugins. More generally, the developer docs are a great place to start understanding how QIIME 2 archives work.
One recommendation I have for you: as you develop a better understanding of how you might implement this, consider reaching out here for feedback on how your types and/or formats might interact with other plugins in the ecosystem. Well-defined concepts can sometimes be useful across multiple plugins, so it’s worth taking the time to consider how they fit in with what already exists, or might exist in the future.
This answer was pretty high-level (as was your question) - feel free to post more specific questions here as they come up.
Best,
Chris 
2 Likes