Let's talk
Contact us

Q&A: Model informed compound selection in drug discovery: The Why and How

Dr. Yurong Lai, Gilead Sciences

Q: Once you use +/- rifampin in cynos to get the appropriate SFs for active T and passive permeability CL from your in vitro systems, can you then use the same SF for passive CL for other compounds, rather than having to run a monkey PK for each compound? In other words, do you still see compound specific SF for passive CL from your in vitro system?
A: No, SFs remain compound specific. Although the Passive permeability can be the same, but it contribution to the PK curve could be different, as the liver to plasma ratio will be the determinant of compound diffusion back to the blood steam (therefore the curve shape). The ratio is determined by the sum of hepatic uptake, passive diffusion and metabolic/biliary CL.

Q: In other words, long time of a compound staying in blook may not be a good drug (although this may be considered as a good PK) for moving on based on my previous question/nation. 
A: Agree! It depends on where is the target for your compounds. In my case, plasma concentration Ctrough is the key for compound selection, therefore, long half life is an important consideration. 

Q: How can we utilize deep learning techniques to significantly improve the prediction of human PK profiles from chemical structure alone?
A: I am not aware of any AI application yet, and I wanted to explore more in this space as well.

Q: What are the best practices for predicting human PK from nonclinical data for the broad range of new modality therapies being developed?
A: Interesting questions. My presentation focuses on small molecule with transporter mechanism. I am not sure if there is a common practice yet to predict all modality therapies.

Q: How to improve the performance of PBPK models at earlier stages of drug development where less compound-specific information is available?
A: Many in silico model tools can be use. For example, Polin & Theil or Roger & Roland method can be used to predict Vss and tissue distribution Kp. Empirical values, e.g difference of plasma binding and the liver binding can also be applied when the data are not available yet.

Q: How to account for species differences in intestinal metabolism when making projections of human PK prediction?
A: For iv PK prediction, the intestinal metabolism can be excluded. But the intestinal metabolism needs to be considered for bioavailability and potential drug-drug interactions.

Q: How should the uncertainty in a PK prediction be calculated and expressed?
A: Sensitivity analysis should be done for all uncertainties and variations. The worst-case scenarios should be taken into the consideration.

Q: How can we improve the accuracy of prediction of interindividual variability, including PK in special populations?
A: It needs to further phenotype the contribution of each enzyme and transporter. We have not done much is this space yet. 

Q: For compound B, does the beta phase, where you highlight the longer half-life, cover Cefficacious and if so, can you tolerate the large Cmax/Min range? And what tissue is the target organ?
A: You are right, the Cmax/min ratio is high for compound B (about 3X). Since Ctrough is an important for Cefficacious, you need the safety margins covering the Cmax. 

Q: In my view, the time of an anticancer compound stays in blood circulation could not reflect whether the compound is good PK or not. Instead, the time and concentration of the compound staying in tumor are the key. Is my understanding correct? 
A: Fully agree! Even some anticancer targets e.g Immuno-oncology targets, only need a pulsatile coverage; Under the circumstance, long half live is bad, and then it is critical to use the model approaches to deselect those compounds….

Q: Furthermore, if an anticancer drug could stay in blood circulation very long, this likely means that this drug will be very toxic to normal tissue. Is this understanding correct?
A: Like the answers for the question 2 above. It depends where is the target. Modeling approaches can help you to determine the PK shape, not just overall CL values. 

Q: On page 14, the V is Vss from iv PK?
A: Yes, Vss is the average from all species. 

Q: How to predict CL more accurately if extrahepatic metabolism plays a significant role as well?
A: You need to characterize it and understand the contribution to overall CL. Practically, the human renal CL values can be estimated using animal data through SSS or allometric scaling. The CL value is added to the human PBPK model.

Q: Is the t1/2 underestimated in your case is mainly due to the underestimation of V? 
A: No, I don’t think so. The ECC mixes the CLuptake and CLmet; here, only CLmet is elimination CL, but CLuptake only contributes to the distribution phase (the compound leaves from plasma into the liver, but not from your body).