Benefit of Retraining pK_a Models Studied Using Internally Measured Data

Abstract

The ionization state of drugs influences many pharmaceutical properties such as their solubility, permeability, and biological activity. It is therefore important to understand the structure property relationship for the acid-base dissociation constant pKa during the lead optimization process to make better-informed design decisions. Computational approaches, such as implemented in MoKa, can help with this; however, they often predict with too large error especially for proprietary compounds. In this contribution, we look at how retraining helps to greatly improve prediction error. Using a longitudinal study with data measured over 15 years in a drug discovery environment, we assess the impact of model training on prediction accuracy and look at model degradation over time. Using the MoKa software, we will demonstrate that regular retraining is required to address changes in chemical space leading to model degradation over six to nine months.

Predicting pKa for Small Molecules on Public and In‐house Datasets Using Fast Prediction Methods Combined with Data Fusion

“…In this study, we were interested in the off‐the‐shelf performance of commercially available pK a models and the use of data fusion in order to improve their accuracy. It should be stressed that it has been well established that pK a predictions as well as predictions of other related constants such as logD on internal datasets can be dramatically improved by re‐training the models using the mispredicted molecules and sometimes even a single training compound from the novel chemical space is enough to make the model perform adequately …”

“…As stated before, applying data fusion to pK a prediction was suggested by Avdeef but there is no such study on the subject published to our knowledge. Also, the performance of the models tends to change over time as shown by Novartis researchers so it was also interesting to see how well the updated versions of the pK a prediction tools performed on our data. In addition, the diversity of functional groups in previous studies has often been limited.…”

“…Bayer co‐operated with the software vendor Simulations Plus to produce the S+pKa model that is commercially available . In a related effort, Gedeck and colleagues at Novartis described their approach of retraining internal pK a models using the commercially available MoKa suite . Both of these models rely on thousands of data points that have been measured inside a single company.…”

Topoisomerase Inhibitors Addressing Fluoroquinolone Resistance in Gram-Negative Bacteria

“…pK a Measurements: pKa values were determined using the UV-metric method as described previously. 101 UV-metric ionization constants were determined on the commercial Spectral Gradient Analyzer (SGA) or T3 instrument (Sirius Analytical Ltd., sirius-analytical.com) as described by Allen et al 102 Test compounds were diluted to 0.04 mM in a cosolvent mixture and titrated three times in 20-40% wt methanol. The titrations were performed at 25 °C and 0.15 M ionic strength, from pH 2 to 12 or 12 to 2 (with delta pH of 0.2) depending on the acidic or basic nature of the test compound.…”

Evaluation of the GastroPlus™ Advanced Compartmental and Transit (ACAT) Model in Early Discovery

“…The pKa values are used to calculate the absorption scaling factors and the solubility versus pH profile. The pKa model was assessed over a large number of Novartis compounds and the median error was found to be 0.7 (24).…”