In vitro compound evaluation using human-derived neural cells is beginning to incorporate microphysiological systems (MPS). Neural MPS includes not only microfluidic devices but has also recently recognized neural organoids as viable MPS platforms. The history of neural MPS utilizing microfluidic devices is extensive, with the development of models that control the positioning of cell bodies and neurite outgrowth, as well as models that mimic neuronal projections through the connection of heterogeneous cell types. This paper presents examples of predicting peripheral neuropathy through machine learning applied to images of cell bodies and neurites in microfluidic devices, as well as the construction of a motor neuron-skeletal muscle model. Additionally, it discusses the responses to contraindicated drugs in Dravet syndrome using brain organoids that reflect biological brain structures. In drug discovery applications of neural MPS, it is essential to develop and utilize appropriate MPS tailored to specific objectives, ensuring biological relevance and reliability for future advancements.