Abstract
Traditional drug screening methods use monolayer (2D) tumor cell cultures, which lack basic features of tumor complexity. As an alternative, 3D hydrogels have begun to emerge as simple, time-, and cost-saving systems. One of the most promising candidates, synthetic alkoxysilane-PEG (polyethylene glycol)-based hydrogels, are formed by "sol-gel" polymerization in an aqueous medium, which allows control over the incorporated elements. Our aims were to optimize siloxane-PEG hydrogels for three different cell lines of skin origin and utilize these 3D hydrogels as a feasible drug (e.g., daunorubicin) screening assay. A drastic increase in survival and the formation of cellular aggregates (spheroids) could be observed in A2058 melanoma cells, but not in keratinocyte and endothelial cell lines. A deep-learning neural network was trained to recognize and distinguish between the cellular formations and allowed the fast processing of hundreds of microscopic images. We developed an artificial intelligence (AI)-assisted application (https://github.com/enyecz/CancerDetector2), which indicated that, in terms of average area of the spheroids treated with daunorubicin, A2058 melanoma cell 3D aggregates have better survival in a hydrogel containing 15% bis-mono-ethoxysilane-PEG.