[Openchemistry-developers] Expressing interest in a project Project: YAeHMOP as a library

[Greg Landrum]

Hi Diksha,

I can reply for both of these projects since I'm the mentor for either of
them.

*YAeHMOP project:*
Here we're only talking about the computational part of the tool, the
software that's in the tightbind directory in the github repo (
https://github.com/greglandrum/yaehmop/tree/master/tightbind)
This is a codebase that is in C (with a bit of fortran or f2c translated C)
and that currently assumes that it's going to be run from the command line
reading from an input file and writing to an output file. There's a lot of
documentation describing what the toolkit does on the sourceforge page (
http://yaehmop.sourceforge.net/). The idea of the project would be to:
1) Create a library form of the package that exposes the important
functions via a public API. We'll need to figure out what the important
pieces of functionality are there, but it won't be a huge number of
functions.
2) Write a good set of regression tests for that public API
3) Replace as much as possible of the legacy fortran (or f2c translated C)
code.
  3a) The first piece of this will be replacing calls to the fortran matrix
diagonalization code with calls to similar functionality provided by the
eigen library.
  3b) The second part will be to replace the code that calculates the
overlap integrals that are used. This is a fairly hairy bit of code and so
we'll have to figure out if there's a sensible way to rewrite it or if we
should make sure that the C code no longer calls out to things from the f2c
library.

That's already a reasonable chunk of work and would be really useful, but
if there's still time I would suggest that it be spent starting to write a
set of unit tests and/or doing a good cross-platform build system based on
cmake.

A good start if you want to do this project would be to get a copy of the
yaehmop source, compile a copy of it, and demonstrate some basic knowledge
of how it works by writing a couple of regression tests (you can pick a
framework for this) using the input files in the yaehmop/examples directory
as input and the code's newly added ability to be called from another
package (you'll find the commits for this in the changes made last year).

*Fingerprint project*:
Here is an brief (and a bit out of date) overview of what chemical
fingerprints are and what is currently implemented in the RDKit:
http://www.rdkit.org/UGM/2012/Landrum_RDKit_UGM.Fingerprints.Final.pptx.pdf
Googling for "chemical fingerprint" will turn up a bunch of hits that may
also help.
Here the idea is to create a new fingerprinting function for the RDKit that
allows the caller to specify the types of molecular "subgraphs" to be
found, the type of features to be mapped onto those, and the hashing
approach to be used.

A good way to start if you want to do this project would be to clone the
RDKit github repo and get it building on your own, find the C++
fingerprinting code, write a function that generates an RDKit fingerprint
using the connectivity invariant from the Morgan fingerprint, and then
write some good tests for that function.

Best Regards,
-greg





On Mon, Mar 13, 2017 at 1:20 PM, <diksha.godbole at gmail.com> wrote:

>
>
>
>
> Respected Sir/Madam,
>
>                                                 My name is Diksha Godbole
> and I am an undergraduate studying Computer Engineering .With this mail I
> would like to express my profound interest in the Project: YAeHmop as a
> library. I have a good C/C++ programming knowledge  and would like to
> discus more about what undertaking this project would mean and what will be
> expected for me .I have also expressed my interest in a previous mail for
> the project: Create a generalized fingerprinting function, I hope with
> your guidance I can decide which project would best suit my interests and
> skill set.
>
> Looking forward to your guidance ,
>
> Thanking You ,
>
> Diksha Godbole
>
>
>
>
>
> Sent from Mail <https://go.microsoft.com/fwlink/?LinkId=550986> for
> Windows 10
>
>
>
>
>
>
>
> _______________________________________________
> Openchemistry-developers mailing list
> Openchemistry-developers at public.kitware.com
> http://public.kitware.com/mailman/listinfo/openchemistry-developers
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/openchemistry-developers/attachments/20170313/7d69ac18/attachment.html>