Request: LibMCS start or run button

Hi,


is it possible to create a "Run" or "Start" option in the LIBMCS menu as button and text. Whenever the options are changed one has to reopen the sdf file.





Another helpful option would be to let LibMCS start in the current path, because I usually have different work directories with projects and I do not store my files under this "default" user


like home/user under LINUX or myDocuments under WIN.


Or is there an option I can set only for LibMCS to do that?





Thank you


Tobias

Quote:

is it possible to create a "Run" or "Start" option in the LIBMCS menu as button and text. Whenever the options are changed one has to reopen the sdf file.

Sure, you are right, it is a "must have button". Actually, there used to be one but due to some nasty workarounds it stopped working so we removed it. It'll be back soon!

Quote:

Another helpful option would be to let LibMCS start in the current path, because I usually have different work directories with projects and I do not store my files under this "default" user like home/user under LINUX or myDocuments under WIN. Or is there an option I can set only for LibMCS to do that?

We'll consider such user setting/preferences in a later stage of the development when the main algorithm is mature and robust - we are really short of capacity (In the longer term LibMCS will probably be integrated in IJC where user preferences have been supported. - Would you use libmcs as part of IJC or you'd rather stick to a standalone application?, just wondering.)





Regards,


Miklos

Hi Miklos,


yes, LIBMCS in IJC in the chemistry section will be very helpful for comparing datasets in different ways.





But I like LibMCS as a standalone version, its a very powerful diversity set analyzer and the power is the GUI and the possibility to examine and play with the molecules and to see the different groups and to start some external calculations with mview (currently the double click is not possible?).





If I want a statistical analysis of molecules I use binary fingerprints with a selected set of substructures and thanks to PubChem this is now very easy to generate such a subset and thanks to the CDK it is very easy to generate diverse-diverse sets with the deterministic isomer generator and thanks to ChemAxon Reactor one create enumerate reaction libraries or one can use SmiLib to generate combinatorial libraries. Or I use any other chemical or biological property and generate myself a supervised or unsupervised model...





But again, the power is the graphical interface of LibMCS and the manifold options, like sorting in a table, checking the parents and childs - at the end the "human" has to understand the scaffold set of certain small molecule classes and has to built some new models, so its not not only automatically done by the computer.





And if the small problems are cleared there will be no real speed problem, people will use quad cores or dual quads so there will be speed improvements, like with a cheap pre-sorting and splitting of the dataset in the beginning.








Tobias





ps: some old problems in my 0.5 version still exist in the table view, the sorting function only sorts if I press it serveral time up/down. Especially when sorting different properties.

Hi Tobias,





many thanks for all those valueble comments you made. It was interesting and edifying to get some insight into the way how you work by combining tools from various vendors. It appears to me that perhaps a pipelining solution would be the most suitable for your needs. Or am I wrong?





thanks again,





regards


Miklos

Hi Miklos,

Quote:

many thanks for all those valueble comments you made. It was interesting and edifying to get some insight into the way how you work by combining tools from various vendors. It appears to me that perhaps a pipelining solution would be the most suitable for your needs. Or am I wrong?

yes, a colleague of mine built already a Kepler workflow for me


using some ChemAxon and CDK tools. And you are probably


aware of Pipeline-Pilot and Taverna workflows.





The problem with workflows is, they are very powerful on paper, they look very pretty


on Powerpoints and some of them are actually working - but they require a pretty deep insight into


A) the workflow technology itself and


B) all the workflow components itself.





Sofar I am currently only working with the Statistica Dataminer workflow system for statistics


and the Genedata Expressionist workflow system for mass spectrometry based solutions.


This keeps me very focused on certain functionality. Because if I have to learn Kepler, Taverna


and PipelinePilot I am probably designing workflows the whole day.





And here it comes to the third caveat


(C) having many pre-built workflows and working examples (content).





Without (C) it probably will never take-off, there will be some fiddling around here and there


but users can not explore the richness (to use a Web 2.0 word) of the rich workflow program


and it will be still faster to use a well designed GUI like Reactor and Textpad for doing the editing.


So without many many working templates it will be very hard for users


to do something because you shift the responsibilty of programming to the user and having a working API and


good programming knowledge the user could already program himself all thee needed tools.





Having Instant-JChem for out-of-the-box experience and having workflows for the more


complex settings is probably a good thing.





Kind regards


Tobias