Psychometricians have long known that little information can be gleened from a single item. Hence, there is a long-standing tradition in education, psychology, and many other fields to use more than one item to measure a latent trait. For example, a math test will always consist of more than one problem (or the single problem with consist of many parts).
Phenotypic data gathered at the same time as genetic data sometimes contains multiple items that measure different aspects of the same latent construct. However, due to the astronomic number of single nucleotide polymorphisms (SNPs) to test, fast analysis methods are generally preferred with much of the prior research employing regression. Regression is fast, but can only predict a single item at time. Hence, associations with rich phenotypic data cannot be properly investigated.
gwsem contains low-level C/C++ code to permit OpenMx to rapidly read genetic data encoded in U.K. Biobank or plink formats. The association between SNPs and a factor model can be explored.