Abstract Study question Is cervical secretion gene methylation profile associated with embryo implantation outcome, and can be used as a non-invasive biomarker for pregnancy prediction in FET cycles? Summary answer Cervical secretion gene methylation profiles show significant associations with implantation outcomes, and can serve as a non-invasive predictor of successful pregnancy in FET cycles. What is known already Cause of implantation failure remains a black box in reproductive medicine, and the exact mechanism of how endometrial receptivity is still behind the curtain. Epigenetic modifications including DNA methylation regulate gene expression and may alter endometrium receptivity. Endometrial DNA methylation changes vigorously only during the reconstruction of a new endometrium throughout each menstrual cycle, suggesting a unique methylation landscape in each regenerated endometrium. Cervical secretions contain implantation-related cytokines and genetic material from endometrium, and the DNA methylation profiles of cervical secretions is highly correlated with endometrium. Therefore, cervical secretions can be used as non-invasive surrogates for investigating endometrial condition. Study design, size, duration This multi-centered, case-control study enrolled 72 infertile females who entered the FET cycle with at least one high-quality blastocyst available between Aug 2017 and Dec 2022. Embryos assessed by Gardner scoring system with grade 3BB or above were defined as high-quality. Cervical secretions were collected on the day right before embryo transfer, and then DNA methylation profile were assessed. Participants/materials, setting, methods These females aged between 20-45 years-old with regular menstruation cycle (24-35 days) and normal BMI (17.5–32.0 kg/m2). Uterine cavity lesions and hydrosalpinx without treatment, adenomyosis, large uterine myoma that distorted endometrium, and thin endometrium (<7mm) were excluded. The association between the cervical secretion gene methylation profile and pregnancy outcome was evaluated via quantitative methylation-specific PCR (qMSP), methylation array, and genome-wide analysis. Different machine learning algorithms were utilized for differential methylation analysis and implantation prediction. Main results and the role of chance Cervical secretion DNA methylome differed between receptive (pregnant) and non-receptive (non-pregnant) FET cycle with genome-wide methylation array. The accuracy of pregnancy outcome prediction using top 2000 differential methylation probes (DMPs) in promoter region can be as high as 86% in all tested sample and 96.4% in pregnant samples. Bioinformatic analysis of six selected candidate genes related to implantation from literature review showed significant differences in DNA methylation between receptive and non-receptive endometrium. The methylation level of the six candidate genes with eight regions were measured by the qMSP platform and showed varying degrees of differential methylation status between the pregnancy and non-pregnancy groups. The AUCs for a single gene/region ranged from 0.53–0.67 for implantation prediction. Ten different machine learning models were applied and the results showed that the combination of candidate genes’ DNA methylation profiles could differentiate pregnant from non-pregnant samples with an accuracy as high as 86.67% and an AUC of 0.81 by logistic regression model. The best model from different machine learning algorithms using selected DMPs of genes related to window of implantation (WOI) can predict ongoing pregnancy with accuracy rates as high as 86.78% and AUC values up to 0.91. Limitations, reasons for caution The results obtained from a relatively small cohort size, and the timing of sample collection is on the day right before embryo transfer. Further investigation of methylation profiles in late follicular phase along with large sample size with euploid embryo transfer are mandatory for validation and clinical applications. Wider implications of the findings We firstly demonstrated the feasibility of cervical secretion methylation profile as a non-invasive testing for pregnancy prediction. The promising result may contribute to predicting endometrial receptivity for cycles with a favorable endometrium and maximize the successful pregnancy rate of embryo transfer. Trial registration number not applicable