chemPK™ V.2

Human PK Prediction Directly from Chemical Structure

Predicts key human oral and intravenous pharmacokinetic parameters directly from structure.
No in vitro ADME or physicochemical data requirements.
Save money and time by allowing pharmacokinetics to be characterized virtually (no synthesis required).
Provides early stage filter for directing chemistry and prioritizing screening.
Superior approach which uses PBPK model optimized from human clinical (in vivo) data.

A virtual screening tool for predicting human pharmacokinetics from chemical structure alone.

chemPK™ Input Requirements
Chemical structure, e.g., SMILES, mol or sdf. Dosing regimen.

chemPK™ Data Delivery
Predicted pharmacokinetic parameters following oral and intravenous dosing regimens (for single and repeat doses). Predicted time concentrations for plasma (or blood) and all tissues/organs with graphical representation. Predicted time series hepatic metabolism, renal elimation and fecal elimination.

How does it work?

chemPK™ utilizes a KNIME workflow-based approach which executes the processes illustrated in Figure 1.

chemPK - PK prediction from chemical structure

Figure 1
chemPK™ workflow processes.

KNIME can be downloaded easily and for free. Cyprotex can then provide the bespoke KNIME nodes (chemPK™) for the workflow including Generate PK inputs, Run PK simulation and Calculate PK parameters nodes.

The Generate PK inputs node includes calculations of the following parameters:
10 Tissue partition coefficients (Kp) trained on in vivo data Adipose, brain, heart, gut wall, kidney, liver, lung, muscle and skin. Tissue partition coefficient for the remaining tissues and organs that are not explicitly represented in the PBPK model. Rate of renal clearance trained using in vivo plasma and urine profiles following intravenous (bolus and infusion), oral and subcutaneous dosing. Rate of metabolism in the liver trained using intravenous in vivo profiles (bolus and infusion). Rate of absorption from the gastrointestinal tract trained using oral in vivo profiles (rapid release).

The Generate PK inputs node includes calculations of the following parameters:

10 Tissue partition coefficients (Kp) trained on in vivo data

Adipose, brain, heart, gut wall, kidney, liver, lung, muscle and skin.
Tissue partition coefficient for the remaining tissues and organs that are not explicitly represented in the PBPK model.

Rate of renal clearance trained using in vivo plasma and urine profiles following intravenous (bolus and infusion), oral and subcutaneous dosing.
Rate of metabolism in the liver trained using intravenous in vivo profiles (bolus and infusion).
Rate of absorption from the gastrointestinal tract trained using oral in vivo profiles (rapid release).

The Run PK simulation node:
Executes the PBPK model with intravenous and oral dosing including a repeat dose option. Requires the parameters generated by the Generate PK inputs node. Output is in KNIME table (data frame) format that can be processed in any way required, e.g.: Converted to XLS Converted to CSV Plotted using Python, R etc. Written to a database using a connector

The Calculate PK parameters node reports the following predicted PK parameters:
CL, t_1/2, MRT, V, V_ss, AUC_∞ and fraction absorbed to the last timepoint. C_max, t_max and AUC_last for all model tissues/organs.

Data

Data from chemPK™

PK Parameter	MFE^a	Rank
Total clearance	2.54	0.51
Half-life (t_1/2)	2.16	0.52
Mean time a molecule resides in the body (MRT)	2.26	0.47
Volume of distribution (V)	2.08	0.71
Volume of distribution at steady-state conditions (V_ss)	2.03	0.69

Table 1
Intravenous PK parameters for an independent test set of 62 compounds.

^a Mean fold error

PK Parameter	MFE^b	Rank
Highest drug concentration in plasma (C_max)	3.30	0.72
Time at which the highest concentration occurs (t_max)	1.73	0.17
Area under the plasma drug concentration-time curve (AUC_last)	3.46	0.71

Table 2
Oral PK parameters for an independent test set of 35 compounds^a.

^a MW < 450 g/mol
^b Mean fold error

Figure 2
Plot of the predicted volume of distribution versus the observed volume of distribution for a test set of 62 compounds.

We are looking for partners with whom we can further develop the chemPK™ model. If you would like to be involved then please get in touch.

Contact enquiries@cyprotex.com for a demonstration.

chemPK™ V.2

Human PK Prediction Directly from Chemical Structure

How does it work?

Data

Data from chemPK™

Contact us to discuss your ADME Tox issues or request a quote

or fill out the form below: