The versatility of somatosensation arises from heterogenous human dorsal root ganglion (DRG) neurons. The critical information to decipher their functions, i.e., the soma transcriptome, is lacking due to technical difficulties. Here we developed a novel approach to isolate individual human DRG neuron somas for deep RNA sequencing (RNA-seq). On average, >9000 unique genes per neuron were detected, and 16 neuronal types were identified. Cross-species analyses revealed that touch-, cold-, and itch-sensing neuronal types were relatively conserved, while the pain-sensing neurons displayed marked divergence. Soma transcriptomes of human DRG neurons predicted novel functional features, which were confirmed using single-cell in vivo electrophysiological recordings. These results support a close relationship the between physiological properties of human sensory afferents and molecular profiles uncovered by the single-soma RNA-seq dataset. In summary, by conducting single-soma RNA-seq of human DRG neurons, we generated an unprecedented neural atlas for human somatosensation.