python - SVM training set format - how to convert 3D scanned objects (of protein scanning microscope) as proper SVM input -

i use skikit-learn package implement svm , need implement in python following task. system input: 3 dimensional scanning of protein (density in each of coordinates of 256x256x256 pixels scanned sample). desired output: 3d coordinates location of amino acids (which given protein comprised of, concatenated building blocks), , identifying amino acids type (from possible 22 types). suggestion use training set: isolated scanning of known single amino acid different angles, zooming, , resolutions - ("learn" svm) general geometric shape. , so 22 amino acids. , then, analyzing of unknown protein provided chain of amino acids sequence (as text), while demanding output described above (aas coordinates localization, , aas type).

please advise: 1. how convert scanning samples of separated amino acids input training set? 2. how classify output format (trivial numerate 1 22 groups , that's all? 3. skikit-learn package proper environment implement such project?