wavemol team mailing list archive

[Stefano Borini <stefano.borini@xxxxxxxxxxxxxxxx>]

So, let's start.

At the moment, I am writing most of the information on the wiki, at

http://forthescience.org/wiki/MolecularDatabase

Plan for development:

- I will commit very soon a prototype mock database interface which will 
return fake data. This will allow us to design a database interface that 
satisfies our needs. We will replace it with the real thing at a later 
stage. Having this mock data provider will also give us the chance to 
start developing the web interface, so it's important to get the db to 
satisfy our needs. (I will detail them in a later mail)

- We have the problem of the ID to be solved. All the chemists out there 
consider a molecule as a set of atoms connected by bonds, so they can 
use InChiKey. We don't. For this project, the molecule _can_ be a set of 
atoms connected by bonds, but in most frequent cases it is a set of 
atoms at given positions in XYZ space. How to assign an ID to a 
molecule? At the moment, my algo does the following. The ID is composed 
of three parts: AA-<brutepart>-<geometrypart>

For the brute formula part: takes all the atomic numbers of the atoms 
composing the molecule, sort them and join them with a \n.
e.g in H2O takes 1,1,6
      (sort) 1,1,6
      (join with \n) "1\n1\n6"

For the geometry part:
make a list of strings made of four entries, with the following format 
string "%d %15.10f %15.10f %15.10f". Values are filled with the atomic 
number (first marker) and the three XYZ coordinates in the unit bohr.
Join each of these strings with \n. This algo is actually wrong in the 
current implementation as I am not taking into account the possibility 
of different order of the lines.

For both cases, the resulting string is hashed and made human-readable 
with the following algo: compute a hex representation of a CRC32 of the 
string, zero padded on the left side. Convert the hex into actual 
characters and b32encode the obtained data.

This algo is not guaranteed to be perfect. Any proposal accepted, but as 
the first two letters of the ID (the AA) are reserved, we can change it 
later. it also has high incidence of collision, due to the CRC32.

The same molecule, rotated or translated in space, gives different ID's. 
This is mostly intentional, as moving a molecule can have an effect on 
the results. In this sense, two spatially rototranslated molecules are 
different.

If you have any questions, feel free to ask.