Abstract 1817: Association of nuclear morphometrical variables with drug cytotoxicity: A proof of principle using NCI-60 anti-cancer drug screening panel

Abstract The National Cancer Institute’s panel of 60 cancer cell lines (NCI-60) have been extensively used to screen compounds for anti-cancer activity in a wide range of molecular and pharmacologic assays. We completed a detailed analysis of the nuclear morphometric variables of 57 cancer cell lines from this drug-discovery panel using recently developed computerized method that allows simultaneously assess 42 variables related to morphology/texture of the cell nuclei. The goal of this study was to determine whether unique nuclear patterns seen in cancer cells can effectively predict response to anti-cancer therapies. Stained sections prepared from paraffin cell blocks were used to measure each cell nucleus by size (e.g. area, perimeters); by shape (e.g. area box, fractal dimensions, roundness); and by optical density of the pixels and texture (e.g. margination, heterogeneity). The obtained values were then summarized at the cell line level, generating a dataset comprising 84 variables, represented by mean values and coefficients of variation. We filtered the complete set of chemical compounds, tested on the NCI-60 panel, based on sufficient GI50 variability, narrowing the list down to 4,839 compounds. These compounds were further refined using the Benjamini-Hochberg false discovery rate (FDR) method at 1% FDR, based on the p-values from “goodness of fit” tests. This process identified 1,314 compounds where the inclusion of morphometric variables significantly improved the fit of our models. To determine which variables were highly predictive for GI50 in each compound, we performed lasso regression per compound of logged GI50 versus dummied tissue of origin and morphometric variables. We tabulated the proportion of compound models selecting each variable as well as the median relative strength for each variable. This procedure identified 10 promising compound models, in which the top 5 values of relative strength from each of the tested variables emerged. Our primary objective in this analysis was to find associations between certain nuclear morphometric parameters of tumors cells and drug sensitivity. Better defining how to use these previously untapped clues may pave the way for future studies able to predict sensitivity to cancer drugs and to help in the selection of the best anticancer treatments. Additionally, our work may draw new attention to compounds not now used for anticancer treatments, but whose increased cytotoxicity can be linked to abnormal morphometric parameters that indicate possible anti-tumor activity. Citation Format: Geula Klorin, Luz María Rodríguez, Edmond Sabo, Grant Izmirlian, Amnon Amit, Paul Meltzer, Anna Roschke. Association of nuclear morphometrical variables with drug cytotoxicity: A proof of principle using NCI-60 anti-cancer drug screening panel [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 1817.