I think the biggest change in the last 5 years has been the emergence of single cell analysis as something more or less common place and not out of the ordinary.
But in terms of what an average day looks like for a bioinformatician, I think little has changed in the last 5 years. We are still using the same toolchains, on the same types of system, and roughly the same types of data to answer roughtly the same types of question, more or less, as we were 5 years ago.
If one thing has changed, it is perhaps that with the falling cost of sequencing, severly underpowered experiments (such as RNAseq data with no replicates) is less common today that it was 5 years ago. But that doesn't make much of a difference to those working with me, as I never would have let them on a project with such data to start with.
In the last ten years, the big changes have been workflow management systems (like Snakemake or Nextflow) and literate programming paradigms (such as R or Jupyter notebooks). Some things, such as DEG analysis have also settled down - 10 years ago there was some discussion to be had as to the best way to analyise an RNAseq data set for differential expression. Now its the closest thing you get to a standard protocol in bioinformatics.
Ian, All: any comment about AI/LLM/ChatGPT? I feel like a troglodyte I haven't used ChatGPT or similar, but I see a lot people using it or including LLM in their projects proposals. Sure thing there is some overhype there, but there is also some genuine useful substance. It seems to me that for teaching and student assessment LLM is having a substantial impact since ChatGPT and friends can answer traditional exam questions pretty good.
LLMs have mostly been helpful for me in writing and coding. I use chatGPT every day, but all of my stuff can be done without it (it just makes things quicker and easier [and makes me a bit lazier too haha]).
I think all of this is spot-on. Additionally, there has been a massive increase in understanding of good practices vis-à-vis coding. Providing reproducible code, creating containers and pipelines (docker, snakemake/next-flow) that are stand-alone and on top of that publishers requiring code to be provided upon publication. These requirements were not really around 10 years ago but have definitely helped improve the field.
Another poster brought this up but I think there is going to be a non-insignificant increase in low-quality bioinformatics (not in terms of tools but applied analysis) hitting the literature because non-bioinformaticians can now use AI (LLMs especially) to write “OK” code but since they don’t write it they don’t really need to understand it.
Advent of formal bioinformatics training/teaching programs is likely the major change in last 25 years.
Many are/were practicing "bioinformatics" without a formal license (while the term was first used by Paulien Hogeweg and Ben Hesper in 1977, it did not become a formal discipline until late 1990s). Difficult to do now-a-days but not impossible.
Advent of Next generation sequencing was perhaps the biggest impetus, though micro-arrays (and the advent of cDNA sequencing before that) was the start to the explosion of information (and the apparent ease of data generation we routinely see now). New generation will never realize how hard it used to be to get a few hundred bases sequenced (and that is a good thing).
On the 25-year timeframe, the availability of sequenced genomes coupled with web-based analysis tools democratized the field of genomics for bench scientists, especially those of us working in model systems. Experiments that once took years (cloning/sequencing a gene, identifying a mutation) or were impossible (reverse genetics) became routine. Bioinformatics has fully transitioned from a niche discipline to an essential skill for all researchers.
Ian, All: any comment about AI/LLM/ChatGPT? I feel like a troglodyte I haven't used ChatGPT or similar, but I see a lot people using it or including LLM in their projects proposals. Sure thing there is some overhype there, but there is also some genuine useful substance. It seems to me that for teaching and student assessment LLM is having a substantial impact since ChatGPT and friends can answer traditional exam questions pretty good.
I think LLM/AI will change things, and probably pretty soon, but I don't think it has get changed the day to day life of a bioinformatician.
LLMs have mostly been helpful for me in writing and coding. I use chatGPT every day, but all of my stuff can be done without it (it just makes things quicker and easier [and makes me a bit lazier too haha]).
I think all of this is spot-on. Additionally, there has been a massive increase in understanding of good practices vis-à-vis coding. Providing reproducible code, creating containers and pipelines (docker, snakemake/next-flow) that are stand-alone and on top of that publishers requiring code to be provided upon publication. These requirements were not really around 10 years ago but have definitely helped improve the field.
Another poster brought this up but I think there is going to be a non-insignificant increase in low-quality bioinformatics (not in terms of tools but applied analysis) hitting the literature because non-bioinformaticians can now use AI (LLMs especially) to write “OK” code but since they don’t write it they don’t really need to understand it.