Background: Prior knowledge of Maximum Tolerated Dose and Dose Limiting Toxicity (DLT) for individual drugs may play a critical role in designing an efficient phase 1 trial of the corresponding combination. However, MTD and DLT information are scattered in published literature and does not in any existing databases. Methods: We recruited various background experts to build up a data curation team for curating MTD prior knowledge in published literature. In the literature screening stage, we combined key term search and active learning based NLP technology to retrieve as many relevant publications. Then, we developed a user-friendly curation tool to fully capturing relevant information from literature. In our complete curation scheme , the data curators were required to capture trial related information(e.g., inclusion retiarii, study design deails), Dose relaterd information (dose levels and MTD) and DLT related information (DLT definition and reported DLT and frequency). Results: Currently, we have curated more than 1300 cancer drug phase 1 trial results. These trials involved both individual drug and drug combination trials. All the curated data is in a structured format. We provided this data set on Drug Combo web application (https://drugcombo.shinyapps.io/DrugCombo/). Conclusion: This expert curated MTD data set from published literature could help physicians and biostatisticians design more efficient drug combination phase 1 trials.
