chemTarget™

Predicts biological target interaction directly from chemical structure

Predicts binding affinity, inhibition constants or other measures of interaction with biological targets, directly from chemical structure.
Uses Cyprotex’s unique pattern recognition software to build models from existing data sets (provided by the customer or from the literature).
Analyzes approximately 10,000 descriptors using linear, random forest, neural network and nearest neighbor methods.
Provides clinically relevant binding/inhibition/activation when used in combination with the pharmacokinetic predictor, chemPK™.
Provides an early-stage filter for directing chemistry and prioritizing screening.

A virtual screening tool for predicting in vitro biological target interaction from chemical structure alone.

chemTarget™ Input Requirements
Chemical structure, e.g., SMILES, mol or sdf. Locations of in vivo target expression. Existing target interaction data (e.g. IC₅₀, AC₅₀ or K_i).

chemTarget™ Data Delivery
Predicted target interaction. Predicted engagement in vivo for specified dose-regimen(s) (minimum, maximum, average) if used in combination with chemPK™.

How does it work?

Drug interaction with a target in vivo depends on strength of interaction with the target and concentration of the drug at the target binding site. Models produced by chemTarget™ predict the strength of interactions, whilst chemPK™ can be used to predict drug concentrations in organs and tissues. Together, these technologies enable structure-based screening of in vivo target engagement.

chemTarget - receptor interaction prediction from chemical structure

Figure 1
Schematic illustrating how chemTarget can be integrated with chemPK™ to predict clinically relevant biological target binding affinity.

Data

Data from chemTarget™

In silico recepter interaction prediction

RMSE*	0.68
R²	0.76
Spearman rank correlation coefficient	0.85

*RMSE = root mean square error

Figure 2
Prediction of JNK3 binding affinity from chemical structure. Results are from 10 repeats of 10-fold cross-validation for a set of 697 compounds.

JNK3 is a potential therapeutic target for several neurodegenerative disorders. chemTarget™ predicts JNK3 inhibition directly from structure with a repeated cross-validation R² of 0.76 for a set of 697 compounds.

In silico receptor interaction prediction

RMSE*	0.64
R²	0.72
Spearman rank correlation coefficient	0.85

*RMSE = root mean square error

Figure 3
Prediction of MK2 binding affinity from chemical structure. Results are from 10 repeats of 10-fold cross-validaton using a set of 670 compounds.

MK2 (mitogen-activated protein kinase (MAPK)-activated protein kinase 2) is a potential therapeutic target in inflammatory disease. chemTarget™ predicts MK2 inhibition directly from structure with a repeated cross-validation R² of 0.72 for a set of 670 compounds.

Contact enquiries@cyprotex.com to discuss your project.