We use a non-linear model, termed ACME,
that reflects a parsimonious biological model for
allelic contributions of cis-acting eQTLs.
With non-linear least-squares algorithm we
estimate maximum likelihood parameters. The ACME model
provides interpretable effect size estimates and
p-values with well controlled Type-I error.
Includes both R and (much faster) C implementations.
For more details see Palowitch et al. (2017)