Application of Evolutionary Computation to Antibody Quantity Estimation in Virtual Crossmatching for Kidney Transplantation – Formulation as an Under-Constrained Problem for Multiple Variables

Abstract

In recent years, the effectiveness of Virtual Crossmatch (VXM) has been evaluated for assessing antibody reactivity to donor HLA in kidney transplantation. In VXM, the impact of recipient antibodies on the donor organ is estimated based on fluorescence intensity (MFI values) obtained from the interaction between HLA antigens―modeled on beads mimicking donor leukocytes―and antibodies. While MFI values indicate the quantity of antibodies against a given HLA, in reality, they represent the sum of antibody levels targeting eplet structures within the antigen. However, current analytical methods do not separate antibody quantities specific to eplet units that bind to HLA and react with antibodies. To address this, we propose a strategy that applies Evolutionary Algorithms to obtain estimated values that simultaneously satisfy constraints based on both least-square error and maximum error, while preserving solution diversity within the search space.