How to hire a bioinformatician

Posted 3 months ago | Originally written on 29 Nov 2023

I recently received a request from a friend to advise him on how to hire a bioinformatician. Having been through an interview process quite recently, I thought I'd put this here as a record of what I believe is a useful approach to hire at the right skill level.

By 'bioinformatician' I specifically mean 'sequencing bioinformatician'. I say this because structural biology also has 'structural bioinformaticians'. If that is the case, then here is what I would do.  

First, they would need to do a technical test involving some coding. Nothing fancy but just to validate that they have the technical skills required to get the work done. This means that ideally they would need to know:

  • Linux and shell commands hopefully with some shell scripting experience,
  • knowledge on how to not only install genomics software but also how to install from source (they often need this),
  • how to script in a mainstream programming language (Python, Ruby, Java),
  • how to do statistical analyses using something like R.

Knowing how to work with databases and SQL is usually an added bonus but for someone with the right technical skills is relatively easy to pick up. 

Next, they would need to know the genomics workflow all the way from quality analysis/control, then specific skills depending on the outcome. For example, for quantitative analyses then they need to map reads, measure, apply a gene model, do differential analysis (DE) for genes or transcripts. For assembly (not my area), then they need to know how to build the genome then measure its quality etc. The same applies to any functional genomics pipeline like ChIP-seq, ribo-seq etc. But the standard is usually just DE. Each workflow will usually also have a set of data resources required in additional to the software such as genomes, gene models, genome browsers etc. which they need to be familiar with how to source and use. 

Finally, they will need some visualisation skills. Typically, R and Python have everything they need so this can be part of the tech test.  

I've also noticed a growing need to have skills on how to make workflows using tools like Nextflow, Snakemake or CWL. These are usually coupled with using compute clusters. Not really necessary but if the infrastructure exists (which most people would not have at home!) then training is usually provided by the institution/company. For example, I had to learn how to use SLURM at my workplace and this training was provided by on site admins. Experience with such tools is great as it can greatly facilitate quick turnarounds. 

I would keep each part very simple. Candidates are usually quite nervous and you can easily lose good candidates due to nerves. They may know the details pretty well but being observed can panic them. Whatever you do, please do not give them a code sample to start off with: this is a huge throw off because different programmers have different styles. Much better is that they write things from scratch then a discussion can follow on how it can be improved. If you can have an experienced bioinformatician administer the test that would then benchmark the candidate against the bioinformatician interviewer's skill. In any event, most people will not have had much experience apart from at MSc or PhD level but in a work environment they will have the potential to grow and get better with time. This approach is also good to enable women and minorities stand a fighting chance. Of course the criteria depend on the hiring budget: the higher the pay the more experience one would be expected to have and the more demanding the test will need to be - but nothing fiendish.