Organic materials are used for an increasing number of applications. The modeling of their properties has certain specificities compared to other classes of materials. In particular, the structural formula of the molecules that constitute them plays a particularly important role, often predominant compared to the role of 3D conformation or crystal packing. As a result, developing predictive models for the properties of such materials most often amounts to finding quantitative structure-property relationshipsQuantitative Structure-Property Relationship (QSPR), where the molecular structure is numerically represented by so-called descriptors. On the other hand, phase changes and chemical properties are of special significance. Depending on the relative weight put on physics and data, modeling approaches range from first-principles simulations to machine learningMachine Learning (ML) techniques. Special emphasis is put on semi-empirical methods as they allow to take the most from available data and physical understanding.