Organic Semiconductors · foundry-ml
foundry_osdb_v1-1
Tabular dataset of zeolite-guest poses and binding energies for machine learning property prediction and screening.
What foundry_osdb_v1-1 is
This dataset provides computational data generated to investigate the ab initio control of zeolite synthesis and intergrowth through high-throughput simulations. Developed by Daniel Schwalbe-Koda and Rafael Gómez-Bombarelli in 2021 and distributed via Foundry-ML and the Materials Data Facility, it catalogues organic structure-directing agents (OSDAs) docked into crystalline zeolite hosts.
Formulated for machine learning in materials science and chemistry, the collection supports tabular regression and classification tasks. Researchers can train models to evaluate host-guest affinity, select suitable organic directing agents, and analyse energetic competition across different framework topologies.
How it was made
The data is structured as tabular records describing docked host-guest poses. Structural inputs include the International Zeolite Association (IZA) zeolite framework code, unit cell crystal lattice matrices, atomic coordinates and numbers (nxyz), as well as the guest molecule's SMILES string, InChIKey, and chemical formula. Spatial descriptors include the OSDA molecular volume and principal component axes, alongside the unit cell loading level.
Target values consist of simulated thermodynamic metrics. These include binding energies normalised per SiO2 unit and per OSDA molecule, directivity energies, host competition energies, and templating energy computed at 400 K. Synthetic accessibility is quantified using the Synthetic Complexity Score (SCScore), and an indicator records whether a given host-guest pair is documented in existing literature.
Key facts
- Authors
- Daniel Schwalbe-Koda, Rafael Gómez-Bombarelli
- Year
- 2021
- DOI
- 10.18126/c5z9-zej7
- Publisher
- Materials Data Facility / foundry-ml
- Modality
- Tabular, text
- File format
- Parquet
- Task categories
- Tabular regression, tabular classification
- Splits
- train
What you can use it for
- Binding Energy Prediction
- Train regression algorithms to predict host-guest binding energies normalised per SiO2 or per OSDA from chemical and crystallographic descriptors.
- Zeolite Directivity and Competition Screening
- Model directivity and host competition energies to screen candidate molecules that preferentially template a specific zeolite framework over alternative phases.
- Structure-Directing Agent Feasibility Analysis
- Assess the synthetic feasibility and templating efficacy of proposed organic molecules using SCScore, geometric dimensions, and 400 K templating energies.
- Literature Synthesis Classification
- Construct classification models to predict whether a particular zeolite-guest pairing has been validated in experimental synthesis literature.
Explore the data
The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from foundry-ml/foundry_osdb_v1-1 at the moment you asked. Nothing is cached or stored — every row above came from that request.
Strengths and limits
Good at
- Integrates full 3D crystallographic features, such as lattice matrices and atomic coordinates, directly with 1D/2D molecular identifiers.
- Supplies multiple thermodynamic targets including binding, directivity, competition, and temperature-specific templating energies.
- Incorporates synthetic accessibility metrics alongside geometric and energy parameters for holistic screening.
Watch out for
- Contains only a predefined training split, requiring users to manually partition validation and test sets.
- Templating energy evaluations are provided only at a single specified temperature of 400 K.
- Field definitions in the reference metadata contain truncated descriptions for several structural and energetic attributes.
Questions
Can I look inside foundry_osdb_v1-1 without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser, with full-text search and a SQL filter. Nothing is downloaded to your machine.
What licence is foundry_osdb_v1-1 under?
Its authors declare cc-by-4.0. Read the terms on the original page before reusing the data.
What energetic targets are available for regression?
The dataset includes binding energies normalised by SiO2 and OSDA, directivity energies, host competition energies, and templating energy evaluated at 400 K.
What molecular representations are provided for the organic guests?
Guest molecules are represented by their SMILES strings, InChIKeys, chemical formulae, 3D atomic coordinates (nxyz), molecular volume, and principal component axes.
What is the purpose of the SCScore field?
The SCScore provides a Synthetic Complexity Score to evaluate the synthetic accessibility of the organic structure-directing agent.
How to cite it
Schwalbe-Koda, Daniel and Gómez-Bombarelli, Rafael. Data for: Ab initio control of zeolite synthesis and intergrowth with high-throughput simulations. Materials Data Facility, 2021. doi:10.18126/c5z9-zej7
Written from foundry_osdb_v1-1's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.