AI-Powered Chemical Language Model Predicts Dual-Target Compounds for Drug Discovery

AI Revolutionizes Drug Discovery with Dual-Target Compound Prediction

Researchers at the University of Bonn have developed an innovative AI system that predicts chemical compounds capable of targeting two proteins simultaneously, potentially revolutionizing drug discovery for complex diseases like cancer 1

1

2

2

3

3

4

4

. This breakthrough, published in Cell Reports Physical Science, utilizes a chemical language model akin to ChatGPT, but specifically designed for molecular structures.

The Power of Dual-Target Compounds

In pharmaceutical research, compounds with multi-target activity are highly sought after due to their polypharmacology. Prof. Dr. Jürgen Bajorath, who heads the AI in Life Sciences area at the Lamarr Institute, explains, "Because compounds with desirable multi-target activity influence several intracellular processes and signaling pathways at the same time, they are often particularly effective - such as in the fight against cancer" 1

1

2

2

.

While co-administration of different drugs can achieve similar effects, it often leads to unwanted drug-drug interactions and varying breakdown rates in the body. The ability to design a single compound with predefined dual effects could overcome these challenges.

Training the Chemical Language Model

The researchers trained their AI model using pairs of SMILES strings, which represent organic molecules and their structures as sequences of letters and symbols. Sanjana Srinivasan from Bajorath's research group elaborates, "One of the strings described a molecule that we know only acts against one target protein. The other represented a compound that, in addition to this protein, also influences a second target protein" 1

1

2

2

3

3

.

The model was fed with over 70,000 of these pairs, allowing it to acquire implicit knowledge of how normal active compounds differed from those with dual effects. This training enabled the AI to suggest molecules that could act against multiple target proteins when given a compound targeting a single protein.

Fine-Tuning for Diverse Targets

To enhance the AI's capabilities, the researchers implemented a fine-tuning phase. This step prepared the model to suggest compounds that could influence different classes of enzymes or receptors, broadening its application in pharmaceutical research. After fine-tuning, the model successfully produced molecules known to act against desired combinations of target proteins 1

1

2

2

3

3

4

4

.

Implications for Drug Discovery

While the immediate creation of new compounds surpassing existing pharmaceuticals may not be the primary outcome, the true strength of this approach lies in its ability to generate novel ideas. Bajorath notes, "It is more interesting, from my point of view, that the AI often suggests chemical structures that most chemists would not even think of right away. To a certain extent, it triggers 'out of the box' ideas and comes up with original solutions that can lead to new design hypotheses and approaches" 1

1

2

2

3

3

4

4

.

This AI-driven approach to drug discovery opens up new possibilities for creating more effective medications with fewer side effects. By predicting dual-target compounds, it could significantly accelerate the development of treatments for complex diseases and potentially revolutionize the pharmaceutical industry.