[en] Highlights: ► Dose- and time-dependent cellular responses are used to evaluate the cytotoxicity. ► The CI can reflect the cell number, cell viability, morphological change, etc. ► The CSVID can capture the dynamic information after cells exposed to toxins. ► The multi-class classification can distinguish the compounds using multi-doses. ► The majority vote strategy (fingerprint) can improve the classification accuracy. - Abstract: An early determination of toxicant compounds of water contaminations can gain critical time to protect citizens’ health and save substantial amounts of medical costs. To determine toxins in real time, a multi-dose classification algorithm using cellular state variable identification (CSVID) is developed in this paper. First, the dynamic cytotoxicity response profiles of living cells are measured using a real-time cell electronic sensing (RT-CES) system. Changes in cell number expressed as cell index (CI) are recorded on-line as time series. Then CSVID, which reflects the cell killing, cell lysis and certain cellular pathological changes, is extracted from those dynamic cellular responses. Finally, a support vector machine (SVM) algorithm based on CSVID is employed to classify chemical compounds and determine their analogous cellular response pathway. In order to increase the classification accuracy, a majority vote of the class labels is also proposed. Several validation studies demonstrate that CSVID-based classification algorithm has great potential in distinguishing the cytotoxicity response of the cells in the presence of toxins.