Getting a scientific prize for open-source software
gael-varoquaux.info
gael-varoquaux.info
I want to see it spread beyond computer programming libraries into areas where sharing is harder, like open source scientific equipment and fully reproducible methods in chemistry experiments.
The ACM (academic computer science group) has awarded prizes to these open source systems projects amongst others recently:
- Wireshark
- Jupyter
- GCC
- Mach
- Coq
- LLVM
- Eclipse
- make
- Java
- Tcl/Tk
The linked article is written like the academy never works on or recognises open source software or implementation work, or using open licences is unusual. That's not true.
In traditional (= non-CS) academia, proprietary software is still very much the norm, and as long as institutes get a free license for academic usage they also don't seem to care about open source too much. I don't know how much precedent there is, but such recognition from traditional academia still seems to be pretty rare and worthy of highlighting.
Top-tier open source libraries for cheminformatics (or other natural science -informatics flavours) would already be a welcome start.
You can put text documents on GitHub describing process, in the same way as you can code and data. If you have some setup with a femtosecond mid-infrared laser or prep requiring a synthesis lab with all the reagents, labware, and safety equipment you can open source the bill of parts, the build instructions and the lab book. It'd probably be very valuable to do that so please do!
Note: I am not in any way affiliated with this research or the labs involved. This came out of a quick search.
https://pubs.acs.org/doi/suppl/10.1021/jacs.6b13031/suppl_fi...
Yes, but that's documented in lab books and procedure documents. Or at least it should be! If it isn't, how are they able to explain their own research? And those can be open sourced.
Even so, very few people will be able to replicate that work outside without a very well funded laboratory or collaboration of their own. Lab notebooks contain a vast amount of tangential or irrelevant data which are distilled into the publication. What good is a process document for obtaining an x-ray structure if the diffractometer costs a fortune and is a shared departmental or even national resource? In your example, how deep does the bill of materials go? Is it sufficient to state that one needs a Bruker FTIR or Coherent optical parametric oscillator or do those have to be decomposed into the lowest-level components?
That's not a reason to not be open about it!
> What good is a process document for obtaining an x-ray structure if the diffractometer costs a fortune and is a shared departmental or even national resource?
I think it's inherently good! Even if you think can't use it right now, it's good to put it all out there for people looking and into the archives to keep it for the future.
> In your example, how deep does the bill of materials go?
Well if bills of materials are available for your components themselves then you don't need to break them down yourself.
We're talking past each other. Openness in methods has been around in the physical and biological sciences for a very long time.
> Well if bills of materials are available for your components themselves then you don't need to break them down yourself.
Reverse engineering a piece of purchased equipment to publish its BoM now becomes a required part of scientific publication? Or, expecting a manufacturer to provide it and authorize it for general release? I don't think that's realistic.
You're listing a lot of reasons why you think it's not worth it!
> Reverse engineering a piece of purchased equipment to publish its BoM now becomes a required part of scientific publication?
No, you mis-read me. I said if it's an existing piece of equipment just say you used that piece of equipment.
> You're listing a lot of reasons why you think it's not worth it!
No, I am saying that it is not a new concept. I am all for it and tried to hew to that standard in the papers I've written.
And since I can't reply at the correct level, I'll take that as a hint that the site doesn't really want this and stop here.
Thanks for the discussion.
Context: I do research in computational chemistry, and write an open source library for this, that could be used for cheminformatics too. I don't really know what is needed for this though, since I never touched cheminformatics.
Maybe I'm just not deep enough into it, but from my impression so far especially when it comes to application-level software (in contrast to specialized research), OEChem and similar closed source libraries seem to be the most widely used ones, with nothing quite comparable available.
Context: Software Engineer that is also currently a biochemistry undergraduate.
Somehow the same for me, this is part of why I started my own project (http://chemfiles.org). I have the impression that for cheminformatics you want to see molecules as graphes, is this true or is a list of bonds enough for usual purposes?
I have heard of OEChem but never used it. I'll try to find some documentation to have a look.
Yeah, that was my thinking.
I've also seen your work on lumol, so you seem to be one of the few people working in the field with Rust! I just recently started writing a SMILES parser in Rust[0], as a first step towards an in-memory graph representation of molecules. I have a first rough draft of that locally, though it's very rough and changing a lot, as I have to adjust it weekly as I'm basically learning the required theory at the same time :D
https://orbi.uliege.be/bitstream/2268/154357/1/paper.pdf (2013)
https://amueller.github.io/papers/wo_the_machine_varoquaux.p... (2015)