OrgFlow: Generative Modeling of Organic Crystal Structures from Molecular Graphs

Mohammadmahdi Vahediahmar; Matthew A. McDonald; Feng Liu

cond-mat.mtrl-sciFeb 22, 2026

OrgFlow: Generative Modeling of Organic Crystal Structures from Molecular Graphs

Authors: Mohammadmahdi Vahediahmar, Matthew A. McDonald, Feng Liu

Organizations: Department of Computer Science, Drexel University · Department of Chemical and Biological Engineering, Drexel University

Abstract

Crystal structure prediction is a long-standing challenge in materials science, with most data-driven methods developed for inorganic systems. This leaves an important gap for organic crystals, which are central to pharmaceuticals, polymers, and functional materials, but present unique challenges, such as larger unit cells and strict chemical connectivity. We introduce a flow-matching model for predicting organic crystal structures directly from molecular graphs. The architecture integrates molecular connectivity with periodic boundary conditions while preserving the symmetries of crystalline systems. A bond-aware loss guides the model toward realistic local chemistry by enforcing distributions of bond lengths and connectivity. To support reliable and efficient training, we built a curated dataset of organic crystals, along with a preprocessing pipeline that precomputes bonds and edges, substantially reducing computational overhead during both training and inference. Experiments show that our method achieves a Match Rate more than 10 times higher than existing baselines while requiring fewer sampling steps for inference. These results establish generative modeling as a practical and scalable framework for organic crystal structure prediction.

Figures & tables

Figure 1: Overview of OrgFlow. OrgFlow learns to generate organic crystal structures from molecular graphs. It preserves covalent bonds through a bond-aware loss, overcoming limitations of inorganic-focused models. The result is accurate, periodic crystals with more than 10× higher match rates and 25× fewer sampling steps.

Panel of Figure 2 — Figure 2: Overview of OrgFlow. 2(a) shows the architecture combining molecular embeddings, periodic edges, and symmetry-preserving layers. 2(b) illustrates the conditional flow-matching process used to generate full crystal structures.

	Drug-like	Small-molecules	Smallest-molecules	Non-drug-like
Learning Rate	0.0003	0.0001	0.0001	0.0003
Weight Decay	0.0001	0.0001	0.0001	0.0
$\lambda_{f}$ (Frac Coords)	600	400	400	600
$\lambda_{l}$ (Lattice)	1	1	1	1
$\lambda_{b}$ (Bond Length)	0.002	0.0015	0.001	0.001
Anneal f	True	True	True	True

Table 1: Training hyperparameters for different organic crystal datasets.

Dataset	FlowMM [ 16 ]	OrgFlow	Increase(%)
Whole dataset	$0.4$	$\mathbf{10.7}$	$+10.3$
Drug-like	$0.1$	$\mathbf{21.94}$	$+21.84$
Not Drug-like	$0.15$	$\mathbf{15.43}$	$+15.28$
Small-molecule	$0.43$	$\mathbf{17}$	$+16.57$
Smallest-molecule	$0.5$	$\mathbf{24.55}$	$+24.05$

Table 2: Match Rate across dataset splits.

Panel of Figure 3 — Figure 3: Examples of predicted crystal structures (green) compared to ground truth (gray). The figures show two-dimensional projections of three-dimensional structures. Only heavy atoms are shown for clarity.

Panel of Figure 4 — Figure 4: Inference efficiency on Small Molecules. OrgFlow saturates at about $20$ steps, while FlowMM needs around $500$ to approach its best Match Rate.

$\lambda_{b}$	Match Rate (%)	Std
0.0	6.7	0.15
5 $\times$ 10 -4	8.2	0.18
1 $\times$ 10 -3	8.4	0.20
2 $\times$ 10 -3	8.5	0.22
5 $\times$ 10 -3	10.1	0.25

Table 3: Effect of bond loss coefficient

\lambda_{b}

on Match Rate.

$\lambda_{f}$	Match Rate (%)	Std
200	8.7	0.21
400	9.8	0.19
600	10.1	0.22
800	11.7	0.13

Table 4: Effect of fractional coordinate weight

\lambda_{f}

(

\lambda_{b}=1\times 10^{-3}

\lambda_{l}=1

$\lambda_{l}$	Match Rate (%)	Std
0.5	9.6	0.16
1.0	10.1	0.15
2.0	11.0	0.18
4.0	13.6	0.21
6.0	6.7	0.12

Table 5: Effect of lattice weight

\lambda_{l}

(

\lambda_{b}=1\times 10^{-3}

\lambda_{f}=600

Crystal Structure Prediction

Crystal structure prediction seeks stable atomic arrangements from composition, molecular identity, or experimental measurements.

See all 32 papers

MolCrystalFlow: Molecular Crystal Structure Prediction via Flow Matching
Feb 17, 2026Cheng Zeng, Harry W. Sullivan, Thomas Egg +8Crystal Structure Prediction Flow-Based Generative Modeling
Department of Chemistry, University of Florida, Gainesville, FL 32611, USA · Quantum Theory Project, University of Florida, Gainesville, FL 32611, USA · Department of Chemical Engineering and Material Science, University of Minnesota - Twin Cities, Minneapolis, MN 55455, USA +6
Fast Organic Crystal Structure Prediction with Unit Cell Flow Matching
Jun 2, 2026Alston Lo, Luka Mucko, Austin H. Cheng +4Flow Matching Crystal Structure Prediction
MIT CSAIL, Cambridge, MA, USA · University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia · Department of Chemistry, University of Toronto, Toronto, ON, Canada +8
Riemannian Flow Models with Reinforcement Learning for Molecular Crystal Structure Prediction
Sep 30, 2026Thomas Egg, Harry Winston Sullivan, Maya M. Martirossyan +8Crystal Structure Prediction Crystal Structure Generation
Center for Soft Matter Research, Department of Physics, New York University, New York 10003, USA · Simons Center for Computational Physical Chemistry, Department of Chemistry, New York University, New York 10003, USA · Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, MN 55455, USA +7

Feb 17, 2026J/K move · Enter open · S save

MolCrystalFlow: Molecular Crystal Structure Prediction via Flow Matching

Cheng Zeng, Harry W. Sullivan, Thomas Egg, Maya M. Martirossyan+7

Department of Chemistry, University of Florida, Gainesville, FL 32611, USA · Quantum Theory Project, University of Florida, Gainesville, FL 32611, USA · Department of Chemical Engineering and Material Science, University of Minnesota - Twin Cities, Minneapolis, MN 55455, USA+6

Crystal Structure Prediction Flow-Based Generative Modeling Flow Matching

Open paper AskPDF arXiv

Molecular crystal structure prediction represents a grand challenge in computational chemistry due to large sizes of constituent molecules and complex intra- and intermolecular interactions. While generative modeling has revolutionized structure discovery for molecules, inorganic solids, and metal-organic frameworks, extending such approaches to fully periodic molecular crystals is still elusive. Here, we present MolCrystalFlow, a flow-based generative model for molecular crystal structure prediction. The framework disentangles intramolecular complexity from intermolecular packing by embedding molecules as rigid bodies and jointly learning the lattice matrix, molecular orientations, and centroid positions. Centroids and orientations are represented on their native Riemannian manifolds, allowing geodesic flow construction and graph neural network operations that respects geometric symmetries. We benchmark our model against a state-of-the-art generative model (MOFFlow) for large-size periodic crystals and a rule-based structure generation method (Genarris) on two open-source molecular crystal datasets. MolCrystalFlow outperforms MOFFlow while achieving competitive performance against Genarris. We also demonstrate an integration of MolCrystalFlow model with universal machine learning potential to accelerate molecular crystal structure prediction, paving the way for data-driven generative discovery of molecular crystals.

Abstract

Crystal Structure Prediction

Explore similar work

MolCrystalFlow: Molecular Crystal Structure Prediction via Flow Matching

Fast Organic Crystal Structure Prediction with Unit Cell Flow Matching

Riemannian Flow Models with Reinforcement Learning for Molecular Crystal Structure Prediction