Application of predictive modeling approaches is able solve the problem of the missing data. There are a lot of studies that investigate the effects of missing values on qualitative or quantitative modeling, but only few publications have been<br>discussing it in case of applications to nanotechnology related data. Current project aimed at the development of multi-nano-read-across modeling technique that helps in predicting the toxicity of different species: bacteria, algae, protozoa, and mammalian cell lines. In this study, the experimental toxicity for 184 metal- and silica oxides (30 unique chemical types) nanoparticles from 15 experimental datasets was analyzed. A hybrid quantitative multi-nano-read-across approach that combines interspecies correlation analysis and self-organizing map analysis was developed. At the first step, hidden patterns of toxicity among the nanoparticles were identified using a combination of methods. Then the developed model that based on categorization of metal oxide nanoparticles’ toxicity outcomes was evaluated by means of combination of supervised and unsupervised machine learning techniques to find underlying factors responsible for toxicity.
