The methodology for classifying human muscle fibers by contraction speed is presented, based solely on transcriptomic data and without the use of classical morphological methods. The input data consisted of CAGE transcriptomic profiles, which allow precise determination of expression levels and transcription start sites in the promoter. To estimate the proportions of cellular components, the MuSiC deconvolution method was applied, using single-nucleus muscle sequencing data from the Heart Cell Atlas project. Based on the obtained estimates a binary threshold for the proportion of fast muscle fibers (10 percent) was defined, demonstrating stable characteristics (AUC = 0.934 and 0.828 for two annotation schemes). Further analysis showed that fiber composition and associated expression profiles differ across anatomical muscle groups. These differences formed the basis for functional annotation, which revealed enrichment for biological processes related to development, specialization of muscle tissue, and possible associations with pathology. The method provides a quantitative, automated, and reproducible assessment of the spectrum of skeletal muscle speed phenotypes, opening the way to standardizing transcriptomic profiling in fundamental and applied research.