Hello Hashim and other future GSoC candidates,

I am a former GSoC student. I worked on the 3D structure generation of Open
Babel in GSoC 2018 and 2019.
https://summerofcode.withgoogle.com/archive/2018/projects/5957928301363200/
https://summerofcode.withgoogle.com/archive/2019/projects/5300703310905344/
Here is our paper based on the GSoC project:
https://jcheminf.biomedcentral.com/articles/10.1186/s13321-019-0372-5

I have some works left related to my project.
As I cannot join GSoC anymore, I hope someone would solve the following
problems.

1. Stereochemistry:
I found out something is wrong
with OBDistanceGeometry::CheckStereoConstraints(), but I still don't
understand what is wrong.
I posted some problematic cases before.
https://sourceforge.net/p/openbabel/mailman/message/36699127/
<https://sourceforge.net/p/openbabel/mailman/message/36699127/The>
The function is replaced now, but it is still problematic.
It compares canonical SMILES of input and output, but when I generate
canonical SMILES from generated 3D structures, SMILES sometimes don't
match.
We need to fix this problem.

If you can list problematic test cases, it would help us to solve the issue.
I think this is an easy task and a good starting point to start
contribution. You can create an issue to report problematic cases.

Here are related codes and discussions:
https://github.com/openbabel/openbabel/pull/1875#issuecomment-547242242
https://github.com/openbabel/openbabel/blob/6e80d6548e04af1dc13f9ff975f18714c7024617/src/distgeom.cpp#L912
https://github.com/openbabel/openbabel/blob/6e80d6548e04af1dc13f9ff975f18714c7024617/test/testdistgeom.py#L24

You can use our test dataset at
https://github.com/n-yoshikawa/ob-fragment-generation/tree/master/data

2. Improve performance:
I wrote distance geometry last year, but it was so slow compared to RDKit's.
If you can improve the performance, it is great.
The slowest part was error minimization in my evaluation.
https://github.com/openbabel/openbabel/blob/6e80d6548e04af1dc13f9ff975f18714c7024617/src/distgeom.cpp#L1171
I doubt initial geometry is strange due to loose distance constraints, but
I am not so sure.

3. Large ring:
The default fragment-based coordinate generation works poorly for unknown
large ring fragments.
Geoff attempted to solve this, but it is not complete yet.
https://github.com/openbabel/openbabel/pull/2019
Using distance geometry for unknown fragment is another possible solution,
but the performance was poor in my implementation.

I used a lot of technical terms here. Please let me know if you have any
questions.

Naruki

2020年2月13日(木) 3:17 Hashim Chaudhry <hashimchaudr...@gmail.com>:

> Hello Devs,
>
> I'm really looking to contribute to Openbabel however, I'm having a hard
> time figuring out where to start. I'd really appreciate if someone could
> point me in the right direction.
>
> *Why I want to contribute?*
> I came across an Openbabel GSOC project idea (Develop a validation and
> standardization filter) at Openchemistry.org and it piqued my interest.
> I'm well aware that the organizations haven't been announced yet,
> nevertheless, I'm looking to contribute to Openbabel to get a start in
> Open-Source work and get familiar with the code-base before GSOC.
>
> *About Me*
> I'm a third-year Computer Science undergrad student. I have completed a
> number of self learning courses for Programming, Data Structures,
> Machine Learning and Data Science(with Python).
> I believe these courses have laid a foundation that would really help me
> to learn and contribute to Openbabel.
>
> Moreover, here's a link to my Kaggle profile
> <https://www.kaggle.com/hashimchaudry>. I'm a beginner finding my way
> into Data Science. It ain't much, but it's honest work.
>
> Regards,
> Hashim
> ReplyReply to allForward
> _______________________________________________
> OpenBabel-Devel mailing list
> OpenBabel-Devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/openbabel-devel
>
_______________________________________________
OpenBabel-Devel mailing list
OpenBabel-Devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/openbabel-devel

Reply via email to