The person who ran the experiment should run the analysis

A sequencing run finishes on a Tuesday. The FASTQ files land on a shared drive. Then they wait, sometimes for days, sometimes for a month, until someone with the right skills has a gap in their queue. In most labs that someone is a bioinformatician who supports several groups at once, and who is very good at their job.

The wait is the obvious cost. The less obvious one is that every judgement call in the analysis moves away from the person who knows the experiment best.

The calls that matter are biological

Which samples were run on a different day. Which replicate looked odd under the microscope. Whether a batch of reagent was off. None of that lives in the metadata sheet, and all of it changes how you read a result. Whether to include a batch in the model, whether to drop an outlying replicate, which threshold to report: these are decisions about the experiment, not about the software. When the analysis is done somewhere else, they get made without that knowledge, or get made late, in a meeting, after the plots are already drawn.

The question is not whether biologists can learn to code. It is whether they should have to, to look at their own data.

What changes when you can run it yourself

You see the quality report the day the run finishes, so a failed library is caught while there is still time to prepare it again. You try the analysis with and without the odd replicate and look at both. You bring a figure to lab meeting instead of a status update, and the questions it raises get answered that week rather than the next month.

What happens to the bioinformatician

None of this removes them. It changes what they spend their time on: building and checking the pipelines a lab relies on, once, and reviewing the analyses that matter, rather than rerunning the same steps for each person who needs a result. That is better use of a scarce skill, and most bioinformaticians we speak to would rather do it.

What it takes

• The steps a core facility would run, with sensible defaults already set.
• A way to change a parameter without editing a configuration file.
• Somewhere to run it that is not a cluster account you have to apply for.
• A record of what was run, so the result can be checked by someone else later.

That list is the brief OpenBionode is built to. To see what it looks like in practice, read your first differential expression analysis. If the list describes the gap in your lab, we would like to hear from you.