Поиск предикторов токсичности иммунотерапии при помощи дискриминантного анализа

Abstract

Immune-mediated adverse events in therapy with checkpoint inhibitors can be severe and even fatal. The identification of indicators that will allow predicting the toxicity of immunotherapy is an urgent task. In this work, an attempt was made to determine the predictors of immunotherapy toxicity using the discriminant analysis method. The meaning of discriminant analysis is to convert a multidimensional array into a one-dimensional indicator based on training samples to predict whether observations belong to groups. After the calculations were made by the program, 2 out of 7 analyzed indicators were selected: CPS and the number of lymphocytes before the start of therapy. Both independent indicators included in the prognostic model are directly related to the value of the discriminant function: an increase in both CPS values and the number of lymphocytes leads to an increase in the likelihood of autoimmune manifestations. The sensitivity of the model was 100%, the specificity was 75%. Based on the constructed model, it is possible to create a calculator that allows predicting the risk of developing immune-mediated complications in a particular patient, according to his laboratory data.