New LOLS machine learning approach facilitates molecular conformer search in complex molecules

A new method based on machine learning yields promising results when searching for molecular conformers even in large molecules
Schematic showing molecules

CEST researchers developed a new machine learning approach based on a low-energy latent space (LOLS) and density functional theory (DFT) to search for molecular conformers.

Molecular conformer search is a topic of great importance in computational chemistry, drug design and material science. The challenge is to identify low-energy conformers in the first place. This difficulty arises from the high complexity of search spaces, as well as the computational cost associated with accurate quantum chemical methods. In the past, conformer search would take up considerable time and computational resources.

A person standing by water
Photo showing Xiaomi Guo

To address this challenge, visiting doctoral student Xiaomi Guo, together with other CEST researchers Lincan Fang, Prof. Patrick RinkeDr. Xi Chen, and Prof. Milica Todorovic (University of Turku) explored the possibility of performing the molecular conformer search in a low-dimensional latent space. This method uses a generative model variational auto-encoder (VAE) and biases the VAE towards low-energy molecular configurations to generate more informative data. In this way, the model can effectively learn the low-energy potential surface and hence identify the related molecular conformers. The CEST teams calls their new method low-energy latent space (LOLS) conformer search.

In a recent publication the authors tested this new LOLS procedure on amino acids and peptides with 5–9 searching dimensions. The new results agree well with previous studies. The team found that for small molecules such as cysteine, it is more efficient to sample data in real space; however, LOLS turns out to be more suitable for larger molecules such as peptides. The authors now plan to extend their structure search methods to more complex materials beyond molecules.

This research paper is published in The Journal of Chemical Theory and Computation under

The code used in this work can be found at

For more details contact

Xi Chen

Academy Research Fellow
  • Published:
  • Updated:
URL copied!

Read more news

Picture of Marika Tervahartiala and Kamyar Hasanzadeh after the event in front of the Hybrid Stage area.
Research & Art Published:

Perspectives on open data: Aalto Research Data Uncovered

Marika Tervahartiala and Kamyar Hasanzadeh took the stage to share their insights on opening geographic and visual data.
Research & Art Published:

Alternatives to fossil-based fibres – Helena Sederholm develops sustainable textiles

Today’s outdoor clothing mainly relies on oil-based fibres and chemicals that are harmful for the environment. In her doctoral research, Helena Sederholm develops environmentally friendly water-repellent textiles. She follows the example of nature.
New Update
Research & Art Published:

Important Update: ITICAT2023 Conference Now Goes Hybrid Mode + Extension of Early Bird Registration and Abstract Submission Deadline!

ITICAT2023 Conference Now Goes Hybrid Mode + Extension of Early Bird Registration and Abstract Submission Deadline
Humanitarian architects at field work in Nepal, sitting on rugs and discussing with local people
Research & Art, Studies Published:

Aalto University is launching a World in Transition Programme for resilience of vulnerable communities

New diploma programme seeks solutions to the most pressing problems concerning the human settlements.