Empowering chemical research with comprehensive, high-performance open-access reaction data.
BORDS is an advanced reaction search engine built on Google's Open Reaction Database (ORD). It provides instantaneous access to millions of chemical reactions, significantly improving search speeds compared to traditional platforms.
📌 Built with: Google Cloud Platform (GCP), Elastic Enterprise App Search, React, Python, and Firestore.
📌 Optimized for: Sub-second query performance (vs. 30+ seconds on other platforms).
📌 Designed for: Chemists, researchers, and data scientists working on reaction prediction, synthesis planning, and retrosynthesis.
✅ High-speed search (seconds vs. 30+ sec) – Built on Elastic Enterprise App Search for optimized query performance.
✅ Rich metadata support – Displays reaction conditions, time-course data, and structural representations.
✅ Scalable cloud architecture – Runs on Google Cloud App Engine with autoscaling.
✅ Molecule rendering & visualization – Uses RDKit and Flask to generate reaction images dynamically.
- Google Cloud Platform (GCP) – Fully managed, auto-scalable deployment.
- Google Cloud Firestore – Stores processed reaction data for real-time retrieval.
- Python (Flask, RDKit) – Handles reaction image generation.
- Node.js (Fastify, Elastic App Search) – Enables ultra-fast reaction search queries.
- Elastic Enterprise App Search – High-speed indexing and retrieval of reaction datasets.
- React.js (Material-UI, Framer Motion) – Provides a seamless and modern UI.
- Google Analytics (gtag.js) – Tracks user behavior for performance optimization.
BORDS indexes millions of single-step reactions into Elastic Enterprise App Search, providing instant access to structured reaction data.
Each indexed document includes:
- Reaction Identifiers
- Conditions & Provenance
- Reaction Time & Outcome Data
- Links to External Chemical Databases
BORDS employs a high-performance ETL pipeline to transform Google ORD data into a searchable format.
- Script:
1_to_json.py - Task: Converts Google ORD protocol buffer files (
.pb.gz) into structured JSON format. - Key Actions:
- Extracts reactants, products, conditions, and metadata.
- Uses multiprocessing to handle large-scale datasets.
- Script:
2_retry.py - Task: Prepares reaction data for indexing.
- Key Actions:
- Stringifies nested fields (identifiers, inputs, outcomes).
- Cleans and structures metadata for efficient searching.
- Script:
3_to_firebase.py - Task: Inserts reaction JSON data into Google Cloud Firestore.
- Key Actions:
- Uses batch inserts for high-speed loading (~2000 docs/sec).
- Implements retry logic for failed transactions.
- Script:
4_to_es.py - Task: Indexes reactions into Elastic Enterprise App Search for fast querying.
- Key Actions:
- Pulls from Firestore, transforms records, and indexes in bulk.
- Implements checkpointing to resume failed syncs.
BORDS runs on Google App Engine, with dedicated services for search, rendering, and indexing.
Used for storing raw ORD datasets and reaction metadata.
Optimized multi-region ElasticSearch cluster with high availability.
Achieves 44ms search time on average.
- Method:
POST - Endpoint:
/api/read/search - Description: Performs a structured search on reaction datasets.
- Method:
GET - Endpoint:
/api/generate/render_molecule?mol=<SMILES_STRING>&type=smiles - Description: Generates molecule images using RDKit.
✅ Expand ORD integration – Support for multi-step reactions & retrosynthesis.
✅ Enhance search filters – SMILES-based substructure search.
✅ Develop API access – Allow programmatic reaction retrieval.
This project utilizes data from the Google Open Reaction Database (ORD), licensed under the CC-BY-SA 4.0 License. BORDS is an open-source project licensed under the CC-BY-SA 4.0 License.
📧 Email: support@bordsearch.com
🌍 Live Demo: bordsearch.com
🚀 Experience the fastest reaction search with BORDS today!