Large Image

Abstract

For a vast library of ligand proteins, it is difficult to directly measure the dissociation constants K_D of each ligand with its target receptor. We propose a massively parallel method to estimate K_D values using “cDNA-display molecules”, in which the ligand protein is tagged with a cDNA molecule that represents its genotype. Our approach is based on fundamental equations governing chemical equilibrium and in vitro selection dynamics, and is quite straightforward. First, through a single selection process and subsequent next-generation sequencing (NGS), the “enrichment ratio” ε_s ≡ y_s/x_s for each sequence is calculated, where x_s and y_s are respectively the molar fraction of sequence s for immediately before selection and that for after selection. Next, K_D values for several “reference sequences” chosen from the library are measured, and a calibration curve representing the relationship between ε and K_D is obtained. Then, we can estimate K_D values for other sequences based on their ε values. We applied the method to two different VHH libraries: one consists of local sequence space (Case 1) and the other consists of global sequence space (Case 2). To evaluate the validity of the method, the K_D estimates for 6 - 28 reference sequences were compared with their measured values. As a result, the correlation coefficients between them were 0.98 for Case 1 and 0.79 for Case 2, suggesting that our method is highly effective. Theoretically, large-scale NGS increases the prediction accuracy, and then the method is expected to become more practical.