Background The most common biomarkers of Alzheimer's disease (AD) are amyloid beta (A beta) and tau, detected in cerebrospinal fluid (CSF) or with positron emission tomography imaging. However, these procedures are invasive and expensive, which hamper their availability to the general population. Here, we report a panel of microRNAs (miRNAs) in serum that can predict P-tau/A beta 42 in CSF and readily differentiate AD from other dementias, including vascular dementia (VaD), Parkinson disease dementia (PDD), behavioral variant frontotemporal dementia (bvFTD), and dementia with Lewy body (DLB). Methods RNA samples were extracted from the participant's blood. P-tau/A beta 42 of CSF was examined for diagnostic purposes. A pilot study (controls, 21; AD, 23), followed by second (controls, 216; AD, 190) and third groups (controls, 153; AD, 151), is used to establish and verify a predictive model of P-tau/A beta 42 in CSF. The test is then applied to a fourth group of patients with different dementias (controls, 139; AD,155; amnestic mild cognitive impairment [aMCI], 55; VaD, 51; PDD, 53; bvFTD, 53; DLB, 52) to assess its diagnostic capacity. Results In the pilot study, 29 upregulated and 31 downregulated miRNAs in the AD group were found. In Dataset 2, these miRNAs were then included as independent variables in the linear regression model. A seven-microRNA panel (miR-139-3p, miR-143-3p, miR-146a-5p, miR-485-5p, miR-10a-5P, miR-26b-5p, and miR-451a-5p) accurately predicted values of P-tau/A beta 42 of CSF. In Datasets 3 and 4, by applying the predicted P-tau/A beta 42, the predictive model successfully differentiates AD from controls and VaD, PDD, bvFTD, and DLB. Conclusions This study suggests that the panel of microRNAs is a promising substitute for traditional measurement of P-tau/A beta 42 in CSF as an effective biomarker of AD.