Predicting Chiral Crystallization of Molecules Without Stereocenters
5 affiliations
- 1School of Business Administration, Bar Ilan University, Ramat-Gan 5290002, Israel
- 2Faculty of Engineering, Bar-Ilan University, Ramat Gan 5290002, Israel
- 3Department of Chemistry and The Institute of Nanotechnology, Bar-Ilan University, Ramat-Gan 5290002, Israel
- 4Institute of Chemistry and Center for Nanoscience and Nanotechnology, The Hebrew University of Jerusalem, Jerusalem 9190401, Israel
- 5Department of Natural Sciences, The Open University of Israel, Raanana 4353701, Israel
Abstract
Given an achiral molecule which lacks a stereocenter, will it form chiral crystals? Predicting the crystallization symmetry mode of molecules has long been a central challenge in structural chemistry. Particularly enigmatic has been the observation that achiral molecules can give rise to chiral crystals, assuming any of the 65 Sohncke space groups. While case-specific interpretations have been proposed over the years, no general and reliable predictive model for that phenomenon has emerged. Here, we present an efficient solution to this long-standing challenge. We do so by developing a high-accuracy artificial intelligence-based framework, capable of predicting whether an organic molecule lacking stereocenters will form either chiral or achiral crystals. With a carefully curated dataset of organic molecules devoid of stereocenters, we trained a supervised neural network classifier to predict whether such molecules crystallize in Sohncke or non-Sohncke space groups, while identifying the molecular descriptor patterns that favor each outcome. The model achieved a receiver operating characteristic area under the curve (ROC-AUC) of 0.954 on a balanced held-out test set drawn from the same data source. Analysis of the descriptors that most influence the model’s predictions identifies molecular features that are statistically associated with each outcome. Predictions of the crystallization mode of anthracene and helicene derivatives exemplify these principles and are discussed in detail. Altogether, our predictive model opens new pathways for the rational design of molecules capable of forming chiral crystals and innovative avenues for crystal engineering, tailored for the many applications of chiral crystals, ranging from enantiomorphic drug safety considerations to enantioselective catalysis, and more.
Related
- Predicting Active and Inactive Targets of Small Molecules from ChEMBL Data Using the Polypharmacology Browser PPB4
- A machine learning-based platform for predicting the potential energy of small molecules
- Interference between multiple photoionization pathways in chiral molecules: Converging continuum results in Gaussian bases
- Designing chiral organic small molecules for circularly polarized light detection via group theory
- SALTED: a symmetry-adapted machine-learning program for predicting electron-densities in molecules and materials
BibTeX
@article{tobaly2026predicting,
title = {{Predicting Chiral Crystallization of Molecules Without Stereocenters}},
author = {Lior Tobaly and Zion Elani and Yitzhak Mastai and David Avnir and Inbal Tuvi-Arad},
journal = {Molecules},
year = {2026},
month = nov,
doi = {10.3390/molecules31214013},
url = {https://www.mdpi.com/1420-3049/31/21/4013},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Predicting Chiral Crystallization of Molecules Without Stereocenters
AU - Lior Tobaly
AU - Zion Elani
AU - Yitzhak Mastai
AU - David Avnir
AU - Inbal Tuvi-Arad
PY - 2026
DA - 2026/11/06
JO - Molecules
DO - 10.3390/molecules31214013
UR - https://www.mdpi.com/1420-3049/31/21/4013
AB - Given an achiral molecule which lacks a stereocenter, will it form chiral crystals? Predicting the crystallization symmetry mode of molecules has long been a central challenge in structural chemistry. Particularly enigmatic has been the observation that achiral molecules can give rise to chiral crystals, assuming any of the 65 Sohncke space groups. While case-specific interpretations have been proposed over the years, no general and reliable predictive model for that phenomenon has emerged. Here, we present an efficient solution to this long-standing challenge. We do so by developing a high-accuracy artificial intelligence-based framework, capable of predicting whether an organic molecule lacking stereocenters will form either chiral or achiral crystals. With a carefully curated dataset of organic molecules devoid of stereocenters, we trained a supervised neural network classifier to predict whether such molecules crystallize in Sohncke or non-Sohncke space groups, while identifying the molecular descriptor patterns that favor each outcome. The model achieved a receiver operating characteristic area under the curve (ROC-AUC) of 0.954 on a balanced held-out test set drawn from the same data source. Analysis of the descriptors that most influence the model’s predictions identifies molecular features that are statistically associated with each outcome. Predictions of the crystallization mode of anthracene and helicene derivatives exemplify these principles and are discussed in detail. Altogether, our predictive model opens new pathways for the rational design of molecules capable of forming chiral crystals and innovative avenues for crystal engineering, tailored for the many applications of chiral crystals, ranging from enantiomorphic drug safety considerations to enantioselective catalysis, and more.
SN - 1420-3049
ER -