Possible to find bioinformatician postdoc if I only could use other's programs and run pipeline?

Question

[michealsmith]

Near the end of PhD. My research is about identification of human genetic variation for certain neurological disorder analyzing exome/whole-genome seq data. I also did two-year bench work based on the discovered variations.

• I'm pretty much familiar with Linux/shell system (sed, awk, GNU..).
  • Lots of NGS big-data experiences (Sequencing, hpcc, WGS, exome-seq, RNA-seq, ChIP-seq analysis, BWA/Bowtie, samtools, GATK, SV detection...)
• Certain level of Python and R (But mostly for simple text processing)

However, I NEVER write my own software. I could only run other's programs and establish pipeline. I would define myself as a disease geneticist being able to run bioinformatics softwares with plenty NGS big data experiences. I think I have deep understanding and insight on human genome and disease genetics. So would I get a bioinformatician job if I could only run other's programs?

When I first entered the field I though bioinformatics=CS/programming/data science for biology and medicine. But now I'm feeling I'm really biology/medicine-oriented, bioinformatics=find biology using CS/programming tools.

Also I feel there won't be much space room for my career if I continue run pipeline without running my own. I'm recently reading lots of machine learning and found essentially majority of our programs are based on that, and it's lots of fun and wide-range application. Maybe I should learn ML and brush my programming skill in order to be more competitive?

[Amitm]

I'll take the risk of saying something. Well you are obviously not alone in feeling this predicament. I myself had a traditional biology training (zoology honors) and then changed gears two years into PhD to completely bioinformatic analyses. I would stress the word analyses as I am not developing softwares. But specifically software development skill wouldn't be a magic pill for any bio. research lab. The way I see a bioinformatician is a bridge who has balanced expertise in both the fields and hence can infer 'findings' from data with some inspiration & effort. Having said that, I do identify with this situation of running other people's software. But software is a loose term and I frequently have to design new analysis strategy by testing, tinkering and iterating over what is already present 'out there'. Though these would not be as impressive or portable as most published softwares are. But again I work in a cancer biology lab and as the bioinfo. postdoc, getting the most insight from the data in the least time is the agenda. At the end of the day I feel passionate about the direct link to the society for the stuff we look into. So, you are much in good shape to get a good postdoc pos. Look for projects you feel inspiration for. And learning ML would of course help if you keep looking into high-throughput data

[michealsmith]

I can't agree more with your nice words here. I feel the exactly the same way. "have to design new analysis strategy by testing, tinkering and iterating over what is already present 'out there'". Exactly. Those seemingly-simple stuff like run BWA could take beginners several months to feel comfortable. We need to make what present "out there" fit into our own lab.

[russhh]

... get stats first

Answer 1

[Istvan Albert]

If you can write a simple text processing program in Python or R and you understand the basic data structures (list, dictionary, set) you can easily become a top notch bioinformatician (why would you want to be postdoc when there are so many more much better jobs in bioinformatics?)

There is a staggering number of existing tools and techniques - the challenge is primarily understanding what they do, how they operate then chaining and combining them together. I happen to believe that every single published analysis out there can be solved by chaining up existing software.

I have noted an odd phenomena with good programmers - they like to write their own stuff even when robust methods are available, they would rather create their own. But there are dangers with that, most problems are more difficult than what they realize. It is really hard to write a better feature counting or interval intersection routine. So they end up with a crappy reimplementation that is neither reusable nor correct. Now this is fine while someone are doing their Ph.D, that's how they learn what not to write and learn how to recognize other people's crappy code. But the whole point of that experience is to learn how once they graduate they should write as little code as possible... :-)

in a nutshell don't worry about programming, bioinformatics has surprisingly little to do with that

[michealsmith]

"bioinformatics has surprisingly little to do with that".......Such good saying....Again now I realized bioinformatics mostly = solve biology problem using computer rather than the other way around; and it took the whole PhD for me to understand this.

Btw, "so many more much better jobs in bioinformatics?", would you please explain a little bit more? Like bioinformatics specialist? Computational biologist at facility core? They are not that different from bioinformatician postdocs in my eyes.

Many thanks!