Background: DIPG is the most aggressive brain cancer in children and adolescents with a median survival of 9 months. Surgery is not possible due to the invasive nature of the disease in a critical area of the brain; only radiotherapy has shown a transient palliative effect on disease progression and no adjuvant therapy has proven to increase disease control. Methods: We launched a large randomized biomarker-driven platform trial to compare three targeted therapies (erlotinib, everolimus and dasatinib) in combination with 54 Gy radiotherapy, in newly diagnosed patients with DIPG (NCT02233049). A central pathological review was performed prior to study entry to confirm the diagnosis of DIPG with H3K27me3 loss (with or without H3K27M mutation) and assess biomarkers. The randomization was designed so that a drug could not be allocated if the corresponding biomarker was absent in the tumor (EGFR overexpression for erlotinib, mTOR activation for everolimus, no specific biomarker for dasatinib). Tumors underwent whole genome sequencing and RNAseq to evaluate prognostic and theranostic biomarkers. Primary endpoint was overall survival. Results: Among the 279 DIPG patients screened for the randomized trial between October 2014 and September 2019, 233 were enrolled in 45 centers in 7 countries; among the 46 screen failures, the diagnosis of DIPG could not be confirmed in 23 patients. No biopsy- or treatment-related death was reported. Serious adverse events > = 3 unrelated to disease progression were reported in 40% of patients, 34%, 36% and 46%, in the erlotinib, everolimus and dasatinib arms, respectively (p = 0.32). With a median follow-up of 5.3 years, median overall survival (OS) from the date of randomization was 9.0 (95% confidence interval, 7.4-14.4), 11.3 (10.3-13.4) and 9.4 (7.7-10.7) months for erlotinib, everolimus and dasatinib, respectively (p = 0.45). OS were not statistically different from those of a historical biopsy-proven cohort. The molecular profiling with copy number aberrations (CNA) had a strong impact on survival (log-rank test, p = 0.0001) as shown by the preliminary analysis of the first hundred patients identifying a group of tumors with very poor prognosis with a HR for death of 1.96 (1.07-3.64) characterized by TP53 mutations (p = 3.05e-11) associated with more CNA (p < 0.0001). Conclusions: This trial demonstrated the feasibility and safety of a large randomized trial for DIPG based on biomarker information from a stereotactic biopsy at diagnosis in an international multicenter setting and generated new tumor biological insights to further develop innovative therapies. Clinical trial information: NCT02233049.