Retour liste des projets 2024

Nature-inspired optimization for predicting molecular structure

Ce projet a été attribué.

Encadrants

  • Radu Dragomir (IDS)
  • Emails: dragomir@telecom-paris.fr
  • Bureaux: 5C45

Nombre d'étudiant par instance du projet:

  • Minimum: 2
  • Maximum: 3

Nombre d'instances du projet :

1

Sigles des UE couvertes et/ou Mots-clés :

optimization, computational chemistry, protein folding, evolutionary algorithms

Image

project image

Description du projet :

How to compute the
minimizer of a multidimensional function? Optimization problems
appear in pretty much all domains of science and engineering.



We will focus on a
classical task in computational chemistry and biology: finding the
stable conformation of N interacting atoms [3]. To do so, we minimize
a function of the position of atoms corresponding to their
interaction energy. It is a highly challenging task, as the energy
function presents a complex landscape with a large number of
different local minima. It is at the heart of the famous protein
folding problem.



We propose to
implement and compare numerical optimization methods which are
inspired from nature. For instance, the particle swarm method [1]
mimics the behavior of a cooperating group of animals, such as a bird
flock, to efficiently explore the optimization space. Another class
of methods is that of genetic algorithms [2]. They are inspired by
Darwinian evolution and use random mutations, pairing and selection
to find good candidate solutions.

Objectifs du projet :

  • Implement in Python one or several global optimization methods such as : random search, simulated annealing, particle swarm optimization, genetic algorithm.
  • Understand the method behavior and influence of parameters
  • Compare the performance on simple test functions
  • Apply to molecular conformation problems. The most classical one is that of predicting Lennard-Jones clusters for up to 100 atoms [3,4].

Références bibliographiques:

  1. A gentle introduction to Particle Swarm Optimization. https://machinelearningmastery.com/a-gentle-introduction-to-particle-swarm-optimization/
  2. Thomas Weise, 2009. Global Optimization Algorithms: Theory and Applications.
  3. David J. Wales and Jonathan P. K. Doye, 1997. Global Optimization by Basin-Hopping and the Lowest Energy Structures of Lennard-Jones Clusters Containing up to 110 Atoms. J. Phys. Chem. A.
  4. List of kown solution of Lennard-Jones clusters. http://doye.chem.ox.ac.uk/jon/structures/LJ.html