The vaginal microbiota is closely related to women's vaginal health. Persistent vaginal microbiota disorder can greatly increase the risk of human papillomavirus (HPV) infection and may cause gynecological diseases including squamous intraepithelial lesion (SIL) and cervical cancer. However, little is known about the microbial contribution during HPV-related cervical cancer development. Herein, we applied machine-learning classifiers to a dataset of clinical and multi-omic molecular features from patients for SIL grading. Among all single-analyte classifiers, the accuracy of vaginal human proteins, species abundance based on 16S sequencing and vaginal metabolites was greater than 0.9. In addition, combining multiple clinical and omic analytes gave better classification accuracy than using single one. Features related to bacterial glycogen and D-amino acid metabolism showed higher importance in multi-omic classifiers, which reveals that the vaginal microbiota of high-grade SIL patients are better able to utilize glycogen secreted by host cells and to synthesize D-amino acids, thus causing an inflammatory response compared to that of low-grade SIL patients. This approach is expected to transfer for prediction of SIL progression and to assist in clinical diagnosis and treatment.