The invention relates to the technical field of biological
medicine, and discloses a multi-
modal data analysis method for
drug screening safety evaluation. By setting experiment duration and sampling intervals, multi-
modal data such as genes and metabolites are collected from a
drug treatment group and a control group at multiple time points, dimension
standardization is performed on the two groups of data based on statistical characteristics of the control group, a comprehensive
state vector is constructed, state changes of adjacent time points are utilized to form a discrete time differential component and a
state evolution function, and a
state evolution function is established. The method comprises the following steps: generating two groups of theoretical trajectories through initial state iteration, comparing the theoretical trajectories with actual observation at each time point to evaluate a multi-dimensional error, performing double accumulation in time and state dimensions to obtain a
toxicity index of a
drug treatment group and a baseline
toxicity index of a control group, and calculating a dimensionless safety
score for drug grading according to the relationship between the
toxicity index and the baseline toxicity index. Therefore, the limitation of dependence on a single index or a single time point is overcome, and the multi-
modal data fusion and full-cycle toxicity quantification capabilities are improved.