There are various approaches to evaluate probabilities of mutations, the result of which heavily depends on what kind of mathematical model is used for analysis. To make the obtained result more robust and reliable, multiple statistical analyses from different points of view are used in this paper to find the probability that the mutations of Omicron variant have emerged naturally. To be concrete, the following four approaches are taken: a Poisson test applied to the count of nonsynonymous spike mutations in the Omicron variant compared with those in the conventional variants; a binomial test applied to the counts of nonsynonymous mutations in the spike protein and synonymous mutations in the whole sequence; a binomial test applied to the count of nonsynonymous and synonymous spike mutations based on the spectrum of point mutations observed in the variants of concern; spectrum comparison of 12 kinds of point mutations in the Omicron variant and the other variants. The results of these analyses all show that the mutation pattern of the Omicron variant is extremely unlikely to emerge naturally in humans, suggesting artificial mutagenesis could have been introduced into an early strain of SARS-CoV-2, possibly cultured in transgenic mice or transgenic mouse cell lines under a genetically heterogeneous environment.