Microbiota-based prediction of chronic infections is promising yet not well established. Early childhood caries (ECC) is the most common infection in children. Here we simultaneously tracked microbiota development at plaque and saliva in 50 4-year-old preschoolers for 2 years; children either stayed healthy, transitioned into cariogenesis, or experienced caries exacerbation. Caries onset delayed microbiota development, which is otherwise correlated with aging in healthy children. Both plaque and saliva microbiota are more correlated with changes in ECC severity (dmfs) during onset than progression. By distinguishing between aging- and disease-associated taxa and exploiting the distinct microbiota dynamics between onset and progression, we developed a model, Microbial Indicators of Caries, to diagnose ECC from healthy samples with 70% accuracy and predict, with 81% accuracy, future ECC onsets for samples clinically perceived as healthy. Thus, caries onset in apparently healthy teeth can be predicted using microbiota, when appropriately de-trended for age. Display omitted •Oral microbiota in 50 four-year-old children were tracked for 2 years•Age-dependent microbiota development is perturbed by early childhood caries (ECC) onset•Shifts in microbiota precede manifestation of clinical symptoms of ECC•Microbial Indicators of Caries, when de-trended for age, can predict ECC onset Teng et al. tracked plaque and saliva microbiota of 50 4-year-old children for 2 years. By distinguishing between aging- and disease-associated taxa and exploiting the distinct microbiota dynamics between disease onset and progression, a predictive model, Microbial Indicators of Caries, is proposed as a method to predict future caries onset.