We are excited to announce that the powerful IBM MAMMAL (Molecular Aligned Multi-Modal Architecture and Language) model is now available for free through our public API at curecancerwithai.com. Developed by IBM Research, MAMMAL is a groundbreaking multi-modal biomedical foundation model designed to enhance cancer and drug discovery research. By integrating diverse biological data—proteins, small molecules, and single-cell gene/omics data—MAMMAL offers researchers a unified platform to drive innovative discoveries.

What is MAMMAL?

MAMMAL represents a significant advancement in the field of biomedical modeling. Its name reflects its ability to align molecular data across various modalities, allowing it to make predictions that are crucial for understanding complex biological interactions. With its unique architecture, MAMMAL is capable of handling tasks related to both classification and regression, serving as a versatile tool for researchers in drug discovery and cancer therapeutics.

The model's strength lies in its training on approximately 2 billion biological samples, making it adept at processing and predicting interactions within biological systems. By leveraging a multi-task approach, MAMMAL has been evaluated on 11 diverse drug discovery tasks, achieving state-of-the-art results on 9 of them, which demonstrates its potential to transform how we approach drug discovery and cancer treatment.

Research Behind MAMMAL

The capabilities of MAMMAL are detailed in the recent paper published in npj Drug Discovery (Nature), titled "MAMMAL — Molecular Aligned Multi-Modal Architecture and Language for biomedical discovery". This comprehensive study outlines the model's development, training processes, and the extensive dataset utilized, comprising roughly 2 billion biological samples. The paper emphasizes MAMMAL's performance across 11 drug-discovery tasks, where it set new standards by outperforming existing models in 9 tasks and matching state-of-the-art results in the remaining 2.

This significant achievement highlights the model's ability to contribute meaningful insights into the interactions between drugs and biological targets, which is critical for advancing cancer research and therapeutic interventions.

MAMMAL in the Cure Cancer With AI API

We have integrated MAMMAL's capabilities into our public API, offering three distinct endpoints that allow users to leverage the model's predictions for various applications:

• Protein–Protein Interaction Prediction (PPI)
  Endpoint: /api/v1/mammal/ppi
  This endpoint predicts interactions between two proteins. By providing two amino-acid sequences (designated as protein_a and protein_b), users will receive a binding-affinity class label of either "1" (indicating the proteins are interacting) or "0" (indicating no interaction). This information is crucial for understanding molecular pathways and designing targeted therapies.
• Drug–Target Interaction Prediction (DTI)
  Endpoint: /api/v1/mammal/dti
  This endpoint allows users to assess the interaction between a drug and its target protein. By submitting a target protein's amino-acid sequence (denoted as target_seq) along with the drug in SMILES notation (denoted as drug_seq), the API will return a predicted pKd value (−log₁₀ Kd), which indicates the strength of the predicted binding—a higher score signifies a stronger predicted interaction.
• ClinTox Clinical-Trial Toxicity Prediction
  Endpoint: /api/v1/mammal/clintox
  This endpoint predicts the toxicity of compounds during clinical trials. By inputting a compound in SMILES notation (denoted as smiles), users will receive a toxicity prediction, where a result of "1" indicates the compound is toxic or likely to fail trials, and "0" suggests it is not toxic. Additionally, a raw score is provided to assist in evaluating the compound's safety profile.

Each of these endpoints is accessible through our API, making it easier than ever for researchers and developers to harness the power of MAMMAL. The API processes requests through POST methods, and while it may take up to approximately 60 seconds to return results due to the model being CPU-bound, the predictions provided can offer valuable insights for ongoing research.

Start Using It for Free

We invite you to take advantage of this amazing opportunity to utilize MAMMAL's capabilities at no cost. Simply create your free API key at /api-keys and explore the documentation for all endpoints, parameters, and code samples at /developers. The free tier allows for 100 requests per hour per API key, making it accessible for both individual researchers and large teams.

Conclusion

MAMMAL's cutting-edge predictions are now accessible for free on curecancerwithai.com. This powerful tool has the potential to significantly enhance cancer research and drug discovery efforts. However, it is essential to note that the predictions generated by MAMMAL are research signals and should not be considered medical or clinical advice. We encourage all users to utilize this resource responsibly and in conjunction with other research methodologies.

To dive deeper into practical AI-for-cancer-research updates, explore our latest blog posts, learn more about our mission, and see how you can support ongoing work on our donations page.

Cure Cancer With AI is an educational research and information platform. It does not provide medical advice, diagnosis, or treatment recommendations; always discuss care decisions with a qualified healthcare professional.