MicroRNAs (miRNAs) are emerging new biomarkers for many human diseases. To fully employ miRNAs as biomarkers for clinical diagnosis, it is most desirable to accurately determine the expression patterns of miRNAs. The optimum miRNA profiling method would feature 1) highest sensitivity with a wide dynamic range for accurate expression patterns, 2) supreme specificity to discriminate single nucleotide polymorphisms (SNPs), and 3) simple sensing processes to minimize measurement variation. Here, an ultra-specific detection method of miRNAs with zeptomole sensitivity is reported by applying bi-temperature hybridizations on single-crystalline plasmonic nanowire interstice (PNI) sensors. This method shows near-perfect accuracy of SNPs and a very low detection limit of 100 am (50 zeptomole) without any amplification or labeling steps. Furthermore, multiplex sensing capability and wide dynamic ranges (100 am-100 pm) of this method allows reliable observation of the expression patterns of miRNAs extracted from human tissues. The PNI sensor offers combination of ultra-specificity and zeptomole sensitivity while requiring two steps of hybridization between short oligonucleotides, which could present the best set of features for optimum miRNA sensing method.