The oral squamous cell carcinoma (OSCC) is detrimental to patients’ physical and mental health. The prognosis of OSCC depends on the early diagnosis of OSCC in large populations.
Here, the present study aimed to develop an early diagnostic model based on the relationship between OSCC and oral microbiota.
Overall, 164 samples were collected from 47 OSCC patients and 48 healthy individuals as controls, including saliva, subgingival plaque, the tumor surface, the control side (healthy mucosa), and tumor tissue. Based on 16S rDNA sequencing, data from all the five sites, and salivary samples only, two machine learning models were developed to diagnose OSCC.
The average diagnostic accuracy rates of five sites and saliva were 98.17% and 95.70%, respectively. Cross-validations showed estimated external prediction accuracies of 96.67% and 93.58%, respectively. The false-negative rate was 0%. Besides, it was shown that OSCC could be diagnosed on any one of the five sites. In this model,
This study provided a noninvasive and inexpensive way to diagnose malignancy based on oral microbiota without radiation. Applying machine learning methods in microbiota data to diagnose OSCC constitutes an example of a microbial assistant diagnostic model for other malignancies.