In this talk, I will overview the challenges and approaches of designing drug combination dose-finding trials by classifying them into algorithm-based, model-based and model-assisted designs. Model-assisted designs are defined as a class of designs that utilize a model for efficient decision making, similar to the model-based design, but its rule of dose escalation and de-escalation can be determined before the onset of the trial in a fashion similar to the algorithm-based design. I will introduce a model-assisted design, namely Bayesian optimal interval (BOIN) drug combination design, as a practical approach to designing drug combination dose-finding trials. The BOIN design is transparent, coherent and simple to implement. Easy-to-use and freely available software will be demonstrated for implementing the design.