drc007
commented
8 years ago On 2 Dec 2015, at 00:07, Mat Todd notifications@github.com wrote:
Q: Different Chemical Drawing Packages: what if we’re all using different drawing packages? Luc mentioned that there is already a comprehensive file converter (is this openbabel http://openbabel.org/wiki/Main_Page?). The mol file is the recommended type, which can be read by everything. Can this be included in future ELNs? The idea is to make sure that chemical information, from anyone, can be read? Chemdraw is still heavily embedded in labs, and people feel happy using it to generate camera-ready figures.
A quick comment about Openbabel, (full disclosure I’m a contributor to OpenBabel), it is released under GPL license I think that you are simply calling it as a web service this probably does not cause any issues but I thought I’d mention it.
There are many chemical drawing packages that now have both desktop and web flavours, (even Chemdraw now has a mobile version) so I think users will expect to be able to use the same chemical drawing package for all their needs.
ChemDoodle web components are already moving along these lines https://web.chemdoodle.com https://web.chemdoodle.com/ and there is clear overlap with the sort of things Luc has so nicely demonstrated. Released under GNU GENERAL PUBLIC LICENSE.
Similarly Marvin JS https://www.chemaxon.com/products/marvin/marvin-js/ https://www.chemaxon.com/products/marvin/marvin-js/ offers drawing and querying tools, requires commercial license.
Elemental http://www.dotmatics.com/products/elemental/ http://www.dotmatics.com/products/elemental/ is another javascript based chemical drawing application, free but not sure of licensing
Ketcher http://lifescience.opensource.epam.com/ketcher/ http://lifescience.opensource.epam.com/ketcher/ javascript based drawing package free and open source, I think it requires a server back end to provide some of the functionality. Uses GNU Affero General Public License.
JSME is a javascript port of JME, JSME is released under a BSD license but I’m not sure about JME (which is only free to non-commercial applications) and I’m not sure about the situation with respect to Open Source. No desktop version.
lpatiny
MedChemProf
alintheopen
cdsouthan
mattodd
AlexanderSavelyev
aclarkxyz
Hangout (#363) took place. Thanks to everyone for giving up their time. Timezones are a challenge, but feel free to have franchised discussions separately - but please just report back here any thoughts/ideas for those who can't make them.
(I'm pasting notes here so that anyone can correct, or supplement, or add questions. If branching action items are needed, please create those separately. Once this post remains inactive for more than a week, it can be closed.)
ChemInfo: @lpatiny outlined the many capabilities of the Cheminfo system, and some future plans related to the development of an ELN incorporating many of these features. Overarching idea is that data are stored on a server, while calculations are carried out locally on the client computer - e.g. NMR data can be stored inside an ELN, but processing can be carried out live in the browser. Cheminfo can do a number of things of immediate use for the OSM project, all the way to allowing calculations of inter-atomic distances from rendered 3D models. There is an upper size limit for browser-based activities of 50K molecules, beyond which (for e.g. visualising Chembl) a web server can be used.
Q: Different Chemical Drawing Packages: what if we’re all using different drawing packages? Luc mentioned that there is already a comprehensive file converter (is this openbabel?). The mol file is the recommended type, which can be read by everything. Can this be included in future ELNs? The idea is to make sure that chemical information, from anyone, can be read? Chemdraw is still heavily embedded in labs, and people feel happy using it to generate camera-ready figures.
Q: What is the primary identifier for molecules? @cdsouthan suggested this is the pubchem CID since it can handle complexities such as salts, ee’s. A feature @mattodd wanted to have, which would help the bench chemist, would be for a system to spot when a molecule being worked on was already associated with a pubchem ID. Could this be captured by Luc’s system?
@MedChemProf : The Workflow Problem. At the moment it's essentially impossible to find relevant synthetic chemistry people have carried out. @mattodd: This is a recurrent and major problem in OSM. We urgently need an ELN that is substructure-searchable. That this might be close is the only reason we’ve not carried on with the manual collection of synthetic data, which was impossibly cumbersome (though very useful for writing up papers). @MedChemProf : Need clearly defined workflows nonetheless - when people make molecules, or want to make molecules, what are the steps. (Do I have this right, Chase? We should e.g. do a flowchart, as a how-to but maybe somewhere like the wiki?)
Responsibility for data. The data generated by the OSM consortium are by default CC-BY. They can be anywhere, physically. However, who takes responsibility for maintaining the existence of the datasets? @lpatiny 's view is that the data are kept available through there being many copies. This is likely to be the most effective strategy. But in the case of a researcher working with OSM at University X, using public ELN Y, who accepts responsibility for making the data permanently available? Should this nontrivial problem fall to the parent institution (“As an institution carrying out primary research we guarantee the permanence of the research record”) or should OSM attempt to guarantee this (perhaps through Luc’s strategy of redundancy) by becoming a legal entity in itself and sourcing funds, then using e.g. CLOCKSS? To be discussed further. Mat: There might be other reasons why OSM should become a non-profit in its own right (longevity, ability to raise funds from more diverse sources etc) so we could talk about this later).
Desirable extras for Luc’s system: i) When writing a procedure in English, for the system to understand which molecules are being discussed, i.e. an active link between table of reagents and the procedure. ii) Automatic import of safety-related information from reagent table into the ELN. iii) Auto-search of other resources, e.g. Chemspider synthetic pages, for related chemistry. iv) @drc007 emphasized that though consideration of data location and provenance was important, the most important thing was further development of the front end of the cheminfo system, along the lines that is currently taking place.
Open Science Prize: The pre-application to the Wellcome Trust by Mat, Luc and others for the MolTrakr idea (#361) is for a major undertaking. The full proposal to the WT would be written if the pre-app is approved (people here are very welcome to be part of that if it's of interest, by the way). The open science prize is smaller, and relies on developing a prototype in 2016. There is a heavy emphasis on proposals that "unleash the power of open content and data to advance biomedical research". A possibility discussed at the end of the meeting between Chris Swain, Chris Southan, Luc and Mat involved a variant of the “who’s making what and where?” problem. Could we develop a tool (as part of Luc’s system, or as a component of the MolTrakr platform) that, through “knowing” what chemistry is being undertaken in a med chem project, is able to connect researchers working on the same or related chemistry in real time. If, for example, a Friedel-Crafts reaction is being carried out in Sydney, a flag would go up to say that yesterday a chemist in Indiana, who also has an open lab notebook, was working on the same chemistry earlier that day. This would create a network that is social (because we want to see what other people are doing), but through raw data captured in lab notebooks. There is also the novel emphasis on reactions that are underway, rather than final compounds. It's also unlikely that anything proprietary could be used for this, since openness provides a strength - more open lab notebooks = more people able to work with you to enhance your research. There is an obvious possible "ResearchGate" model here for monetisation, along the lines of "27 people are working on this chemistry today! Sign up for details." though I would not want to pursue this as part of this prize. (Chris Southan said this reminded him of Biostars. Mat: not quite - that’s more of a Stackexchange site for expertise. We’d be talking about something different that helps you find people who are doing things related to what you’re doing, using passive, behind-the-scenes searching rather than Q&A. More like the ads that appear in Gmail, without that uneasy feeling).
Immediate ways forward: Luc’s ELN will be ready for testing at the end of 2015. However, we can already start posting data in the right format for automated indexing and extraction, via posting RXL, mol and JCAMP files. I'll make a separate issue on what’s needed from the bench chemist. Done: #365.
Not discussed in this meeting:
1) How best to store data in the Master Sheet? Strings are important for now, but hopefully generation of molecule-related data/strings is more automated in future.
2) The SGC has a ChemReg system. OSM have been offered a trial. Does anyone have any time to evaluate, as a place to store biological activity data for OSM compounds?
3) (unrelated, but occurred to me as we were talking). The HRMS calculator Luc showed: which molecular formulae best match the HRMS peak found. Could be adapted for elemental analysis calculator? Given % found from elemental analysis, and a suspected molecular formula of the analyte, which solvents/salts/water and in what ratio would provide the best fit to the observed data? This would automate the (sometimes rather humiliating) manual shoehorning we sometimes have to do with elemental analysis data in which we try to make the data fit some realistic combination of molecules.