Well, it's similar to all the approaches which use natural or fractal phenomena as input to generate music: it's essentially the output of a non-optimal random generator arbitrarily mapped to musical features. There are too many arbitrary human-made decicions between the input and the output to claim that it is "music of spheres" or "music of dna or proteins", i.e. the musical effect of any structures in the input data is rather arbitrary.
Your observation on non-optimal random generative mapping misses a crucial point, the key challenge lies in establishing a topological invariant mapping, which ensures that the intrinsic structures and properties of the input data are preserved in the output. This becomes even more complex when dealing with higher-order structures that span multiple modalities.
Yet the topologies of celestial bodies, or RNA or protein structures, nor the patterns generated by cellular automata or other popular sources have nothing to do with music. You could just as well interpret the patterns in a herd of cows or a flock of birds musically. What you ultimately hear as music is the arbitrary, man-made interpretation of a quasi-random set and its mapping onto music. Through this mapping and its active design by a person, it becomes virtually irrelevant which data was used as input. I have worked as a professional musician and producer and also have a PhD in molecular biology and biophysics and have been involved in algorithmically generated music for thirty years. You can take any random number generator and quantize it in a musically optimized way and get similar results. There is nothing characteristic that makes the experienced listener of such music think "oh, that's typical of protein music".