Sample with no characteristic absorption can be identified by refractive index features. In this work, qualitative and quantitative identification of THz spectra of polypeptides using self-organization feature map (SOM) artificial neural network has been demonstrated. The absorption and refractive index features of three polypeptides, including Argreline Acetate, Alarelin Acetate, and Bivalirudin Trifluoroacetate, were measured by using the terahertz time-domain spectroscopy technique in the range 0.2–2.2 THz. The experimental results show that the three measured polypeptides present high similarity in absorption spectra but difference in refractive index spectra. After the network training process, the collected spectra were identified by the well-trained SOM network at another time. Analyzing the result we can see that the refractive index spectra are clustered and identify much better than the THz spectra of polypeptides. The study indicates that refractive index spectra can also be clustered by the SOM artificial neural network for identification of THz spectra especially when there is no obvious difference in absorption but significant difference in refractive index spectra.