Skip To Content

Partitioning (logP or logD)—are you using or measuring the right descriptor?

April 15, 2021
by Sanji Bhal, Director, Marketing & Communications, ACD/Labs


What’s in a single letter? Most of us who’ve worked in chemistry know about logP. The partition coefficient makes it into Lipinski’s rule of five and most post-secondary (if not secondary) educations.

Yet it masks an important subtlety.

Say we let a compound partition to equilibrium between two immiscible solvents. Divide the concentration in one by the concentration in the other, take the log, and we get logP. Simple enough.

But hold on—what do we mean by compound? When it comes to logP, we mean one exact chemical structure. If a compound ionizes, it’s not the same structure. And since most compounds investigated in pharmaceutical and pharmacological research contain ionizable sites, it’s not logP we should be concerned about, but logD.

LogD describes the distribution of all forms of the compound at a specific pH. Un-ionized, ionized, ionized at some other position—all species go into the equation. Unlike logP, which is pH-independent, logD changes with pH as the fraction of each species shifts.

(The natural conclusion is that logP and logD are the same for un-ionizable compounds. But that’s rarely applicable for pharmaceutical R&D.)

I spent years as an organic/medicinal chemist before realizing this, and it turns out I’m not the only one. Many chemists either use the terms interchangeably or incorrectly refer to logD as logP. But that’s no reason to let this state of affairs continue. Pedantry aside, such inaccuracies could cause serious problems if a compound’s properties are miscommunicated, leading people to inaccurate conclusions about its viability.

So let’s drill the point in with an example.

The gastrointestinal (GI) tract has a changing pH environment (Figure 1).


Figure 1 Changing pH environment in the GI tract.

Within each compartment of the GI tract, an ionizable compound like 5-methoxy-2-[1-(piperidin-4-yl)propyl]pyridine would ionize to a different extent. (The molecule has two ionization centers: pyridine with pKa 4.8 and piperidine with pKa 10.9.)

Figure 2 shows the changing dominance of each ionic species with pH.


Figure 2 The changing ionic forms of 5-methoxy-2-[1-(piperidin-4-yl)propyl]pyridine with pH.

Since the compound changes ionic forms, its logD must change as well, as Figure 3 illustrates.


Figure 3 The logD curve of 5-methoxy-2-[1-(piperidin-4-yl)propyl]pyridine.

We see that ionization drastically affects octanol-water partitioning. (Remember that on a log scale, a difference of 3 units is a 1000-fold difference.) So, it would be painting a false picture to simplify 5-methoxy-2-[1-(piperidin-4-yl)propyl]pyridine’s lipophilicity to a single constant.

The conclusion we draw from predicting the logD profile is contradictory to the one we draw from looking at logP alone. In the physiologically relevant pH range (1–8), 5-methoxy-2-[1-(piperidin-4-yl)propyl]pyridine has high aqueous solubility and low lipophilicity, because it exists primarily in its ionized forms. Thus, we would expect membrane permeability to be poor. Whereas logP, which deals with the physiologically non-relevant form, would tell us that membrane permeability is high. (Perhaps, but not within the human GI tract!)

This example is in drug discovery, but the reasoning applies to other fields. For example, environmental chemists might study the behavior of chemicals affected by the pH of different soils or of acid rain. They would be interested in the partitioning of the species that actually exist at that pH, not just in the neutral species that might not even be the dominant form.

Finally, this example uses predicted data to make the point, but the same argument applies to experiments. LogP should be measured when the compound exists in its neutral form (> pH 12 in the case of 5-methoxy-2-[1-(piperidin-4-yl)propyl]pyridine), or logD should be reported at a specific pH.

Remember: when it comes to partitioning, it’s crucial to tell your ‘P’s from your ‘D’s.

Send me more info!

Subscribe to receive more information about Property Prediction from ACD/Labs

This field is for validation purposes and should be left unchanged.


Your email address will not be published. Required fields are marked *

Send me more info!

Subscribe to receive more information about Property Prediction from ACD/Labs

This field is for validation purposes and should be left unchanged.