Lansky Consulting is a small statistical consulting and training company
in Burlington, Vermont, specializing in the design and analysis of
biological assays for pharmaceutical companies and government agencies
worldwide.  A key portion of our consulting involves estimation of
relative potency from cell-culture assays run on 96-well plates.
To meet this demand, we have been building and using a package in R
that implements "modern bioassay analysis" using robotic randomization
with mixed models.  This package makes heavy use of the R packages nlme,
lme4, lattice and grid, among others.

Relative potency estimation entails assessing the equivalence of shape
for dose-response curves in a test-reference pair, and, if the shapes
are similar enough, constraining them to be parallel and estimating
the distance beween them.  The classes of models we use for fitting
dose-response relations include four-parameter logistic multilevel
models as well as linear dose-subset multilevel models for assays which
do not have asymptotes within their dose range.  Using blocked designs
and analyzing the data using multilevel models yields much more precise
estimates of potency than can be attained with single-level models, yet
most packaged software for bioassay analysis currently used in wet laboratories
cannot model blocked designs.  By building our toolkit in R, we also have
considerable flexibility in adapting the modeling framework to additional
variance components such as those due to the strip-plot design, location
effects, or serial dilution error.  We are devoted to pioneering
the use of randomized designs in bioassay, and have R routines that
generate the code needed to drive robotic pipettors to perform randomized
strip-plot experiments in the laboratory.  (Randomized assays cannot
be feasibly performed manually, so most laboratories don't use them,
despite regulatory guidance.)

Our current efforts are towards building, validating, and commercializing
a web-based customized client interface to these utilities.  Clients will
automatically retrieve robot code to perform randomizations, and then
after running the assay, submit their plate readouts for an assay and
obtain results through a simple interface to our rich set of tools
for complex analyses.  This will be provided as a validated service
complete with tracking/auditing capabilities and GxP compliance.
Assay results are reported in concise (one-sheet per assay) yet
intricately informative graphical summaries that are automatically
tailored to the client-specific protocol.  These "assay detail sheets"
have several interconnected graphical elements (built using lattice and
grid) and have already proven to be an impressive element in
presentations to managers and regulators.