Abstract Background and Aims The majority of children with idiopathic nephrotic syndrome (INS) and adults with Focal Segmental Glomerulosclerosis (FSGS) and Minimal Change Disease (MCD) receive glucocorticoid treatment at diagnosis. Only those with a high likelihood of having monogenic disease or a contraindication to steroids might avoid this treatment. About 10% overall will not respond to steroids and we have no reliable way of prospectively identifying these patients. Therefore, some patients will receive a futile treatment which is accompanied by significant side effects. DNA methylation (DNAm) is an epigenetic mechanism meaning that it can induce stable but reversible changes in gene expression without any change in underlying DNA sequence. DNAm has shown great potential as a treatment stratification tool, for example, DNAm data is used in oncology to identify which patients are likely to benefit from alkylating chemotherapy. We investigated whether DNAm can predict initial response to steroids in children and young adults with nephrotic syndrome. Method Three hundred and seventeen patients with INS were selected from the NephroS and NURTuRE cohorts. All patients were diagnosed with INS ≤ 30 years of age and those who underwent a renal biopsy had a histological diagnosis of either FSGS or MCD. Peripheral blood DNAm measurements were generated using the Illumina MethylationEPIC Beadchip (>850, 000 CpG sites). Clinical data was used to label patients by their initial response to steroids (sensitive, n = 156, or resistant, n = 161). Machine learning models were created to predict steroid response from the DNAm data. Models were generated using elastic net following feature filtering, and model hyperparameters were tuned and performance measured within the context of cross validation. To exclude whether cumulative steroid exposure prior to sample collection had impacted our results, the CpG sites in the final model were compared to those identified in a published study examining steroid exposure and DNAm (4). Results The 317 INS patients had a median age at diagnosis of 5 years (IQR 2-10) and a median time between diagnosis and DNAm sample collection of 4 years (IQR 1-10). The steroid resistant group were made up of patients with known monogenic disease (n = 75, 24%) and those without pathogenic variants (n = 86, 27%). Initial response to steroid treatment could be predicted with 65% accuracy and an area under the curve (AUC) of 0.75, (sensitivity 0.65, specificity of 0.66, see Fig. 1) using DNAm levels at 14 CpG sites. There was no overlap between the 14 CpG sites in our prediction model and those that are known to alter with steroid treatment. Conclusion We have demonstrated that peripheral blood cell DNAm profiles are a promising predictor of steroid response in INS. Further work to incorporate genetic data into the prediction models is underway and external validation of the results in a separate cohort of patients is required.