Entering edit mode
8.6 years ago
Crackjack
▴
20
I am curious to know the current status of efforts to standardize the language of lab protocols (which may lead to better reproducibility and easier automation).
A casual search gave a few like
- Antha (alpha version)
- Transcriptic's Autoprotocols ("open" source but quite Transcriptic-specific)
- Emeralds's Symbolic Lab Language (closed source and Emerald-specific)
- BioCoder (defunct?)
Any more?
Standardizing everything is not the right solution. There are areas where I believe standards are useful and others where I think it doesn't help. Protocols are in such an area. We're constantly pushing boundaries in different areas. A standard here is going to play perpetual catch-up and then what happens when some zealots decide to enforce such standard for publication? No more new methods because it can't be described by the standard. I believe anyway that the whole reproducibility issue is ill-defined with different people having different ideas of what reproducibility means; we could use a standard definition of reproducibility ;). The issue is better tackled by properly training scientists and not trying to pretend that a one-page advert in a commercial journal is actually a scientific article.
@Jean-Karim Heriche: Thanks for weighing in!! :) I was hoping to hear the current status of pro-standardization efforts. But let me engage your opinion: (1) The perpetual catch up problem: That is a problem for all standardization efforts for all sub-domains in all developing fields of study. So that cannot be a reason against standardization. (2) The fear of possible enforcement of such lagging standards on newly developed protocols: This too is possibility in every other developing field too, and yet you don't see that happening (there will be huge protests if that ever happens). I don't see any reason to perpetuate ambiguity beyond when it is unavoidable (like in describing new protocols, for example). On defining reproducibility: One crucial aspect of any such definition should address ambiguities in conveying/communicating protocols, don't you think?
Playing catch up makes developing a standard a long term commitment. Because of this it's a factor to consider when deciding how much resources to devote to a particular effort. As for enforcement, I think that not enforcing standards in some ways is detrimental to their uptake. There are only two ways people adopt standards: one is through enforcement and the other is when the standard provides some benefit over the cost of using it. There are very few successes of standard uptake in the computational biology area. Most of these concern data formats and these came about as a combination of enforcement and equipment vendors implementing them for easy export from their machines/software.
As for communicating protocols without ambiguities, this is where proper training comes in with the allowance for actually writing as much details as necessary in a paper. Far too often due to space constraints, protocols are vague, or just a reference to another paper that leads to a broken paper trail. Consider also that sometimes there are factors that you would never consider recording (e.g. the sex of the lab technicians handling the mice for your experiment).
Fair point.
Enforcement of use of standardization for established steps in a protocol causes no problem (it is the enforcement for new steps which can be optional). As for the benefit, @DanielSwan gives the biggest motivation, others being coordinating with cloud labs, CROs etc
Besides not helping with lab automation, this option, I believe, is harder than standardization. Almost everyone already understands the importance of clear communication in research. But the problem is there is no agreement on what is 'necessary detail' and hence the variability among trainers themselves.
The difference here is that these protocol descriptions can interact with automation equipment, so the idea is that you can define your protocol and have it amenable to automation. If you're interested in genuine reproducibility I'd argue automation is essential.
It's also rather the entire underpinning of the scientific method, so any attempt to transfer the increasingly important messaging around analytical reproducibility in bioinformatics (providing a Docker container with your paper) to a wet lab (providing a standardised description of your protocol) is surely welcome. I don't see this as 'zealotry'.
Antha and Transcriptics were the only two I bumped into on a recent survey I must admit.
Kind of frustrating if you've ever seen a bunch of equipment controlled via LabVIEW.
I am currently on a workable solution. I will use a publicly posted programming language agnostic standard that would allow extensions using existing sophisticated tools. If you are interested in such thing or contributing to it please contact me.
Also please be aware that Transcriptic imposes severe restrictions in terms of trying their technology or extending it (you need to sign an agreement with them) . Based on that it's unlikely that anybody in that space that is not a dedicated client will use their software.