Recognition of spatial relations between pairs ofsubexpressions is a key problem of recognition of handwrittenmathematical expressions. Most methods for spatial relation classification are based on handcrafted rules and geometric indicesextracted from the subexpression bounding boxes. In this work,we propose new spatial relation features that combine subexpression bounding box and intra-subexpression information, alongwith prior knowledge about the general position and size ofsymbols. Instead of handcrafting features, we train artificialneural networks to learn the useful features from two kinds ofhistograms. The first type captures the relative positions and sizesof the subexpression bounding boxes. The second captures therelative positions and shape of a pair of symbols, called dominantsymbols, extracted from the main baselines of the evaluatedsubexpressions. We evaluate and compare our features with twostate-of-the-art features on a benchmark dataset. Experimentalresults show that our features obtain better accuracy than thesetwo features.