The invention relates to a multi-
cancer-species
pathological feature
analysis method and
system based on a unified framework, and belongs to the field of
data processing. In the method, a unified multi-
cancer-species analysis framework is constructed, a graph network is introduced to formally characterize the cross-
cancer-species (such as 16 cancers)
pathological feature similarity, and a NOTEARS
algorithm is applied to aggregated
pathological features to carry out causal discovery, so that the cross-cancer-species pathological feature similarity is found. Therefore, integrated analysis of association and
causality of pathological characteristics of
multiple cancer species in a unified space can be realized, isolation of a traditional single cancer species model is overcome, and migration and fusion of knowledge among different cancers are promoted. Meanwhile, the group features similar to the target patient are obtained from the graph network for
causal analysis, the biological reasonability and stability of
causal inference are remarkably improved, the pathological features with the real driving effect can be more reliably identified, and deep analysis of the pathological features is achieved.