To efficiently analyze interaction and function between proteins, the present invention relates to a method for forming a light-induced protein nanocluster, comprising: an expression vector preparation step of preparing a first expression vector including polynucleotides coding a first fusion protein including a light-induced heterodimer-forming protein and a first self-assembly protein, and a second expression vector including polynucleotides coding a couple protein that forms a homodimer with said light-induced heterodimer-forming protein, or a second fusion protein including said couple protein and a second self-assembly protein; a transformed cell, tissue or individual preparation step of transforming cells, tissues or individuals using said first expression vector and second expression vector; and a light radiation step of radiating light having a wavelength for inducing the formation of heterodimer between said light-induced heterodimer-forming protein and said couple protein, to said transformed cells, tissue or individuals.