About BrainRoute

Advancing drug discovery through comprehensive blood-brain barrier permeability predictions

What is BrainRoute?

BrainRoute is a comprehensive platform for predicting and analyzing blood-brain barrier (BBB) permeability of small molecules. Our platform integrates experimental data, computational predictions, and advanced filtering criteria to help researchers identify promising drug candidates for neurological applications.

The platform contains a curated dataset of 3,800+ molecules with computed physicochemical and structural properties, filtered through multiple drug candidacy rules to provide reliable predictions for BBB penetration.

Interpreting Results

BrainRoute predictions and computed properties are intended as research screening aids. They should be reviewed alongside experimental evidence and domain expertise, not treated as definitive ground truth.

Molecule Tags

Tags describe where each molecule came from and how much review it has received on the platform.

Training

Molecules used to train the BrainRoute prediction models.

Predicted

Molecules added after a user prediction from the BrainRoute Streamlit prediction tool.

Verified

Molecules reviewed through the platform verification workflow using user-submitted experimental data.

CNS MPO Score

The CNS MPO score is a multiparameter optimization score that summarizes how CNS-like a molecule is based on properties linked to brain exposure. Higher scores generally suggest a compound has a more favorable balance of physicochemical properties for central nervous system drug discovery.

In BrainRouteDB, the score is calculated from component scores for lipophilicity, molecular weight, topological polar surface area, hydrogen-bond donors, and ionization-related behavior. These components are combined into a single approximate CNS MPO score for quick comparison across molecules.

Important Disclaimer

The original CNS MPO method uses measured or calculated LogD at physiological pH. BrainRouteDB does not currently calculate true LogD for every molecule in the database, so the platform uses available computed descriptors as a practical approximation. For newly predicted Streamlit molecules, LogD may fall back to LogP when a dedicated LogD calculation is unavailable. Treat this score as a screening aid, not as a replacement for experimental CNS exposure or full pharmacokinetic evaluation.

How It Works

1. Data Collection

We aggregate molecular data from existing databases and experimental studies, focusing on compounds with known BBB permeability profiles.

2. Property Calculation

Physicochemical and structural properties are computed for each molecule using the PADEL tool, including molecular weight, logP, polar surface area, and more.

3. Multi-Filter Classification

Molecules are evaluated against five industry-standard drug candidacy rules (Lipinski, Veber, Egan, Ghose, and PAINS) to ensure data quality and drug-likeness.

4. Prediction & Verification

BBB permeability is predicted using machine learning models. Researchers can contribute experimental data to verify predictions and improve model accuracy.

Classification Filters

Physicochemical Properties

Polarity (TPSA)

Topological Polar Surface Area measures the sum of surfaces of polar atoms in a molecule. Important for BBB permeability as highly polar molecules struggle to cross the BBB.

Lipophilicity (LogP)

Logarithm of the partition coefficient between octanol and water. Determines membrane permeability and is critical for BBB penetration.

Structural Properties

Molecular Size

Larger molecules generally have reduced BBB permeability.

Aromaticity

Classification of molecules with aromatic rings vs. non-aromatic, affecting metabolic stability.

Ring Count & Heterocycles

Number of rings and presence of heteroatoms influence drug metabolism and BBB crossing.

Special Classes

Classification of peptide-like and lipid-like molecules which have distinct permeability profiles.

Drug Candidacy Rules

Lipinski's Rule of Five

Predicts oral bioavailability by assessing drug-likeness. Compounds violating rules are less likely to be good drugs.

• Molecular Weight ≤ 500 Da
• LogP ≤ 5
• Hydrogen Bond Donors ≤ 5
• Hydrogen Bond Acceptors ≤ 10

Veber's Rule

Pharmacokinetic guideline predicting oral bioavailability based on molecular flexibility and polarity.

• TPSA ≤ 140 Ų
• Rotatable Bonds ≤ 10

Egan Rule

Predicts blood-brain barrier and human oral bioavailability using lipophilicity and polarity.

• LogP ≤ 5.88
• TPSA ≤ 131 Ų

Ghose Filter

Defines "drug-like" chemical space based on molecular properties commonly found in successful drugs.

• Molecular Weight: 160–480 Da
• LogP: −0.4 to 5.6
• Molar Refractivity: 40–130

PAINS Filters

Pan-Assay Interference compoundS filters identify molecules that frequently produce false positives in biological assays, reducing experimental noise and improving data quality.

Flags structural patterns known to interfere with assay results

Why These Filters?

Quality Control: These filters remove compounds with unfavorable drug-like properties, ensuring the dataset contains molecules with realistic potential for development.

BBB Relevance: Rules like Egan and Veber directly address BBB permeability, making them ideal for neurological drug discovery.

Industry Standard: These criteria are widely used in pharmaceutical research and have decades of validation behind them.

Comprehensive Coverage: Using multiple rules provides complementary perspectives on drug-likeness, reducing bias from any single filter.

Reproducibility: Transparent, rule-based filtering ensures consistency and allows researchers to understand exactly how the dataset was curated.

Platform Features

Advanced Search & Filtering

Filter molecules by physicochemical, structural, and drug-likeness properties.

Detailed Profiles

View comprehensive molecular information including all calculated properties.

BBB Predictions

Machine learning-based predictions for blood-brain barrier permeability.

Community Verification

Submit experimental data to improve predictions and validate results.

Data Export

Download molecular data and predictions for further analysis.

Open Access

All data freely available to support neurological drug discovery.

Team & Contact

BrainRoute is developed by a dedicated team of researchers and developers passionate about advancing CNS drug discovery. Get in touch with us for collaborations, questions, or feedback.

Project Team

Soham Shirolkar

Project Lead, Lead Developer

University of South Florida

📧 sohamshirolkar24@gmail.com

Lewis Tem

Lead Developer

Developer

📧 lewistem8@gmail.com

Leah W. Cerere

Visualization & Documentation

Designer

📧 leahcerere@gmail.com

Noura E. Ahmed

Visualization & Documentation

Designer

📧 nemase00@gmail.com

Olaitan I. Awe

Project Supervision & Co-Lead

Institute for Genomic Medicine Research

📧 laitanawe@gmail.com

Get In Touch

Email Support

Have questions or suggestions? Reach out to our team leads:

Collaborate

Interested in collaborating on research or expanding BrainRoute?

→ Propose a collaboration

GitHub

View the source code, report issues, or contribute:

→ github.com/omicscodeathon/brainroutedb

Get Started

Explore our dataset, submit experimental data, or learn more about how BrainRoute can support your drug discovery research.