Generate molecules under constraint:

safer, more efficient or less toxic? 

Predict properties, odor, toxicity ...

AlchemAI is the ultimate solution.

Here is a shortened and marked path

to reach your goals faster.

identification of the need


Generation of molecules

under constraints

Safer, less toxic, more efficient molecules ... AlchemAI can generate new molecules

Prediction of properties

Odor, toxicity, Melting temperature, enthalpy of formation, solubility, new uses … AlchemAI algorithms can be trained to predict many characteristics


definition of your specifications

train neural networks


AlchemAI uses the principles of QSPR (Quantitative Structure-Property Relationship) to predict a number of characteristics of molecules.

The cause and effect links between structure and property of a molecule being complex, they are approached by Machine or DeepLearning algorithms (neural networks).


The models developed on AlchemAI thus involve 3 essential elements for

training and prediction:

Data base

• public

• owners

(in some cases)

Molecular descriptors
Representation of the molecule with informational content
• Properties (ie density)
• SMILES (structure)
• Data from the

quantum chemistry

QSPR Models

• Machine Learning algorithms
• Neural networks

generate new molecules under constraints


AlchemAI uses AI technologies capable of creating latent spaces (mathematical space with several hundred dimensions containing all possible molecules).

AlchemAI then explores this latent space with genetic algorithms to generate new molecules, predicting their characteristics, and then evaluate them with trained neural networks.

select the best candidates


The molecules generated by AlchemAI will be classified and then selected by combining its expertise in chemistry and that of the client in its field.

The best results should then be tested and analyzed.

This selection resulting from in silico work reduces the number of laboratory tests.