A novel methodology was extended for modeling the detailed composition of petroleum heavy vacuum resid fractions. The resid molecules were organized in terms of basic structural attributes: cores, intercore linkages, and side chains. The identities of the structural attributes were determined both from the extrapolation of chemical characteristics of light petroleum and the analysis of detailed mass spectrometric measurement of heavy resid fragmentation products. A building block library was constructed containing 600 attributes. The molecular composition was constructed by the combination of attributes, or building blocks, into discrete molecules. The quantitative abundance of each molecule was determined by the juxtaposition of a set of structural attribute probability density functions (PDFs) constraining pure hydrocarbon and heteroatom mixtures. Quantitative structure–property relationships (QSPRs) were applied to calculate the bulk properties of both the constructed molecules and the mixture. The adjustable parameters of the PDFs were determined using an optimization loop that employed an objective function that contained a term for each of the available analytical data points. The resulting optimal molecular compositions were in good agreement with the experimental structural information.