SQL vs NoSQL databases for next-gen data?

[uri.laserson]

I am interested in people's best practices in using a database to store sequencing data.

I perform lots of immune sequencing, and each one of my reads gets highly annotated. Early on, when I would want to compute something, I used to iterated through an entire flatfile, building some data structure in python (typically dictionaries of dictionaries). Then I realized that I am doing exactly what databases were designed for, including searching, grouping, etc.

So I elected to try MongoDB, which is a noSQL database. It was easy to use, as you can just dump JSON objects into it, which was a very natural conversion from python. So far, the performance has been very fast, and I am quite pleased.

But I am now preparing to generate a large amount of additional data, and was wondering whether it would make sense to try out a traditional SQL database, like MySQL or Postgres. Any thoughts as to how they all compare for sequencing data? How do people typically organize their data in a relational database?

Thanks!
Uri

[Luyi Tian]

Maybe you should try an ORM(Object Relational Mapper) that mixes the features of Object and relational database. I strongly recommend sqlalchemy, which is written in python and has clear documents and examples.