Hey, I am new to Graph neural networks. I want to know if my approach to solving my problem is okay. I have to predict the property of a chemical compound (regression). As input, I have elements (for e.g N,C,O) for a particular compound. I have input properties of elements as features (such as atomic number, energy, melting points, etc.). But problem is that my compound has not given structure (i.e not graph) it is just a collection of elements.
My initial approach was to consider it as a bipartite heterogeneous graph considering elements and compounds as two different types of nodes and elements node has feature say Nx1 and the compound node has singular value. But I am not confident in this approach .
So what could be the approach here?