Transcript Exploring C-Chem with numeric MM and Ab

Exploring C-Chem with numeric MM
and Ab-Initio methods
Masood Malekghassemi
What the heck is up with this janx?


Comptuational Chemistry Methods

Step 1: Molecular Mechanics

Step 2: Ab-initio quantum chemical methods
Analysis and translation of Ab-initio quantum
chemical methods to molecular mechanics rules

Step 3: Artificial Intelligence, neural network sounds
proper for the application
Molecular Mechanics
Molecular Mechanics

Identities


Rules


The different kinds of atoms and/or groups – may
be particular atoms in specific functional groups
Governs the quantization of energy of particular
shapes and orientations of the molecule's
constituents
Constituents

Atoms and/or groups governed by rules through
their identities
The Rules

The rules are the main difference between this
program and other molecular mechanics
programs

Provides a generic interface to govern a system
through energetic interactions

See MM class hierarchy on slide 3

Can be generated from arbitrary information

Will be made more generic by having types in
themselves be data structures rather than hardcoded enumerations
Molecular Mechanics

Self-Consistent Field Method (SCF)

Iterate over various orientations and shapes,
checking for lower energies.

Maximize energy or minimize energy


Hill-climbing
Portray via file and/or a basic GUI (perhaps using
Irrlicht or Ogre3D, rather than homebrew, to simplify
process)
What My Molecular Mechanics
Section does so far:

Effectively nothing.

It has the rules done

It has the identities done

It has the constituents done

It does not have fully working I/O


I've been debating with myself over the trivial matter of
file format for too long
It does not have the iterative SCF method
implemented yet
Ab-initio Methods


Due to my current concentration on MM, I've
put very little attention to the ab-initio methods.
However, I've already gotten started somewhat
on its coding by making a generalized integrator
that finds the average of a function via point
generator (thus allowing easy implementation of
multiple Monte Carlo methods using the same
interface) and then multiplies it by the area of its
domain
Talk to me in about a month for actual planning
and diagrams
Artificial Intelligence


This is the lofty goal of the project. If I only
finished up to the end of my molecular
mechanics part or ab-initio part I'd be happy.
However... This'd be amazing awesome to get
started on and done with.
Talk to me in about two or three months for
actual planning and diagrams of the AI idea
Conclusions
The gist of it: Talk to me in five to six months, and
I swear to you, I'll have the thing working :-P