Lanterna Rally

Author: Keri Danielle Hannie

Publisher:

Published: 2020

Total Pages: 0

ISBN-13:

The drug development process in the United States is an expensive and lengthy process, usually taking a decade or more to gain approval for a drug candidate. The majority of proposed, early stage therapeutics fail, even though the typical process narrows from hundreds or thousands of small molecules down to one late stage candidate. One reason for failure is due to the drugs poor or unexpected absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties. Researchers attempt to predict ADMET properties as a way to help prioritize compounds for lead development to minimize expense and time. It was the overall goal of this project to further the prediction of two ADMET properties (absorption and distribution) through the development and application of quantitative structure-activity (QSAR) relationship computational models predicting human intestinal absorption (HIA), Caco-2 permeability (in vivo & in vitro measurements of absorption), and protein binding (measurement of distribution). These combined models would then be paired with additional experimental methods to help prioritize compounds for future ligand discovery efforts in our lab group and for our collaborators. Five computational QSAR models for each of these three properties were created using different molecular descriptor types and solvation models in an effort to examine which approach resulted in optimal performance. The model development process and validation stages of these QSAR models is outlined herein, along with analysis and discussion of commonly mispredicted compounds. Performance was similar across all models (independent of the molecular descriptor used and the solvation models applied. Future efforts at model development will depend on the size of the dataset to be analyzed. If the dataset is small, the i3D-Born solvation models will be used because these models better represent physiological conditions and performed slightly better than the other models. However, if the dataset is large, the 2D descriptor models will be used as these models do not require that a time and resource-intensive conformational search be performed and because it performed nearly as well as the i3D-Born solvation models. There were no common structural features consistently found associated with mispredicted structures. As such we are unable, at this time to pinpoint classes of compounds to avoid in future effortsThe experimental methods outlined in this work focused on developing methods to determine protein binding, specifically determining a fast, inexpensive workflow to classify the difference between high and low protein binding small molecules. Two techniques were used to determine protein binding of small molecules to bovine serum albumin (BSA): fluorescence polarization (FP) competition, and Nano Differential Scanning Fluorimetry (NanoDSF). FP assays quantifies the change in polarization of a target fluorophore between its protein bound and free states, an equilibrium that can be impacted by the presence of small molecule competitors. This method can be performed in a quantitative manner, but it also requires more time and more expensive and specialized instrumentation. In contrast, NanoDSF determines the melting temperature of BSA in the presence (higher) or in the absence (lower) small molecules by determining the intrinsic fluorescence of tryptophan and tyrosine residues while applying a temperature gradient. This method is qualitative, at least in our approach, but is very fast and requires much less expensive instrumentation. In our hands both techniques were successful in distinguishing differences between small molecules exhibiting low and high BSA binding. In summary, this project was successful in that we 1) developed computational tools capable of correctly predicting ADMET properties including HIA, Caco-2 permeability, and protein binding and 2) developed experimental workflows to quantitatively and qualitatively separate small molecules into low and high affinity BSA binders. With these in silico models and in vitro methods established, future research in our group and with our collaborators can make use of these tools to help prioritize compounds in ligand/ inhibitor discovery efforts..