A recent model for nuclear architecture stresses the importance of active (energy consuming) processes in determining the large-scale organization of chromatin. For example, a territorial organization of chromosomes in interphase according to chromosome territories is a basic feature of large-scale nuclear architecture. Such architecture is also known to be non-random since gene rich regions often tend to be more centrally located than gene-poor regions, while there is also some evidence for radial positioning of chromosomes by size. The earlier model considered variations of activity arising from gene density, but that model was incapable of incorporating cell type specificity of chromosome organization. The more refined model presented here considers gene expression as a proxy for activity. Using this model, we recover the spatial separation of euchromatin and heterochromatin seen in vivo as well as show that theoretical predictions compare well to experimental results for the distribution functions of gene density and chromosome centre of mass obtained through chromosome painting. We also show that active chromosomes are more rougher and less spherical, that the active and inactive X chromosomes are differentially positioned and that the contact probability for each chromosome follows a power law with a range of exponent values comparable to those seen in experiments.