README.md

metadata

license: cc-by-nc-nd-4.0
tags:
  - chemistry
  - bioinformatics
  - drug-discovery
  - blood-brain-barrier
configs:
  - config_name: classification
    data_files:
      - split: train
        path: classification/train-*
      - split: validation
        path: classification/validation-*
      - split: test
        path: classification/test-*
  - config_name: regression
    data_files:
      - split: train
        path: regression/train-*
      - split: validation
        path: regression/validation-*
      - split: test
        path: regression/test-*
dataset_info:
  - config_name: classification
    features:
      - name: SMILES
        dtype: string
      - name: label
        dtype: int64
      - name: InChI_Key
        dtype: string
      - name: scaffold
        dtype: string
    splits:
      - name: train
        num_bytes: 1040685
        num_examples: 7593
      - name: validation
        num_bytes: 106674
        num_examples: 727
      - name: test
        num_bytes: 136667
        num_examples: 942
    download_size: 515702
    dataset_size: 1284026
  - config_name: regression
    features:
      - name: SMILES
        dtype: string
      - name: label
        dtype: float64
      - name: InChI_Key
        dtype: string
    splits:
      - name: train
        num_bytes: 81036
        num_examples: 963
      - name: validation
        num_bytes: 7613
        num_examples: 84
      - name: test
        num_bytes: 9294
        num_examples: 100
    download_size: 56403
    dataset_size: 97943

BBB Dataset

The paper is under review.

[Github Repo] | [Inference Model] | [Cite]

Abstract

The blood-brain barrier (BBB) restricts most compounds from entering the brain, making BBB permeability prediction crucial for drug discovery. Experimental assays are costly and limited, motivating computational approaches. While machine learning has shown promise, combining chemical descriptors with deep learning embeddings remains underexplored. Here, we introduce TITAN-BBB, a multi-modal architecture that combines tabular, image, and text-based features via attention mechanism. To evaluate, we aggregated multiple literature sources to create the largest BBB permeability dataset to date, enabling robust training for both classification and regression tasks. Our results demonstrate that TITAN-BBB achieves 86.5% of balanced accuracy on classification tasks and 0.436 of mean absolute error for regression. Our approach also outperforms state-of-the-art models in both classification and regression performance, demonstrating the benefits of combining deep and domain-specific representations.

Dataset Details

This dataset is an aggregation of different literature sources (please see the paper to check the references).

Classification Task

The number of samples for BBB- and BBB+ is presented below (corresponding to TABLE I in the paper).

Set Name	BBB+	BBB-
Training	4,564	3,029
Validation	434	293
Test	638	304
Total	5,636	3,626

Regression Task

For the regression task, based on the classification dataset, only compounds with logBB values were utilized. This resulted in a subset with 963 samples for training, 84 samples for validation, and 100 samples for testing.

Dataset Usage

Classification

Use the code below to load the dataset for classification task.

from datasets import load_dataset 

dataset_dict = load_dataset("SaeedLab/BBB", "classification")

Regression

Use the code below to load the dataset for regression task.

from datasets import load_dataset 

dataset_dict = load_dataset("SaeedLab/BBB", "regression")

Citation

The paper is under review. As soon as it is accepted, we will update this section.

License

This model and associated code are released under the CC-BY-NC-ND 4.0 license and may only be used for non-commercial, academic research purposes with proper attribution. Any commercial use, sale, or other monetization of this model and its derivatives, which include models trained on outputs from the model or datasets created from the model, is prohibited and requires prior approval. Downloading the model requires prior registration on Hugging Face and agreeing to the terms of use. By downloading this model, you agree not to distribute, publish or reproduce a copy of the model. If another user within your organization wishes to use the model, they must register as an individual user and agree to comply with the terms of use. Users may not attempt to re-identify the deidentified data used to develop the underlying model. If you are a commercial entity, please contact the corresponding author.

Contact

For any additional questions or comments, contact Fahad Saeed (fsaeed@fiu.edu).