 Obtaining good initial structures is the main challenge for the computational study of transition states. Here, fast and accurate predictions for transition state of gas phase reactions are achieved by machine learning based on inter-atomic distances. This article was authored by Sung Huang Choi.