Neonates who are born preterm (PT) are usually characterized by immature physiological development, and preterm birth (PTB) is the leading cause of neonatal morbidity and mortality if intensive medical care is not available to PTB neonates. Early prediction of a PTB enables medical personnel to make preparations in advance, protecting the neonate from the subsequent health risks. Therefore, many studies have worked on identifying invasive or noninvasive PT biomarkers. In this study, we collected amniocentesis-derived (at the second trimester of gestation) amniotic fluid (AF) samples. At delivery, AF samples were classified into PTB or full-term birth (FTB). We first applied protein mass spectrometry technology to globally screen AF proteins, followed by specific protein validation with ELISA. We identified four protein biomarkers of PTB, including lactotransferrin (LTF), glutathione-disulfide reductase (GSR), myeloperoxidase (MPO) and superoxide dismutase 2 (SOD2). Further analyses demonstrated that their abundances were negatively correlated with neonatal weight and gestational age. In addition, by mimicking survival rate analysis widely used in tumor biology, we found that LTF and SOD2 were prognostic factors of gestational age, with higher levels denoting shorter gestational age. Finally, using the abundances of the four protein biomarkers, we developed a prediction model of PTB with an auROC value of 0.935 (sensitivity = 0.94, specificity = 0.89, p value = 0.0001). This study demonstrated that the abundances of specific proteins in amniotic fluid were not only the prognostic factors of gestational age but also the predictive biomarkers of PTB. These four AF proteins enable identification of PTB early in the second trimester of gestation, facilitating medical intervention to be applied in advance.