TABLE 4.

Invitrodb and tcpl package enhancement summary.

Enhancement	Invitrodb v3.5 and tcpl v2.0	Invitrodb v4.0 and tcpl v3.0
Curve-fitting models	Models included hill, gain-loss, and constant	In addition to hill, gain-loss, and constant, models included Polynomial 1 (Linear), Polynomial 2 (Quadratic), Power, Exponential 2, Exponential 3, Exponential 4, and Exponential 5 based on BMDExpress and encoded by R package dependency tcplFit2
Activity hit calls	Hitcall was binary: 0 = negative, 1 = positive, −1 = Unable to fit (usually due to fewer than 4 concentrations)	Hitcall is continuous as the product of three proportional weights: median response and top of model exceed the cutoff, and AIC is less than the AIC of the constant model fit
Potency estimates	Potency estimates were based on modelled active concentration series, including ACB (activity concentration at baseline, 3bmad), ACC (activity concentration at cutoff), and AC50 (activity concentration at 50% of maximal response)	Based on the program BMDExpress v2.0, tcplfit2 modelling outputs new potency and uncertainty estimates related to a benchmark dose (BMD) as defined by the Benchmark Response (BMR) level
Stand-alone pipelining	In addition to connecting to a tcpl database, tcplLite connection would create flat files structured like invitrodb for stand-alone pipelining applications	tcplLite is no longer supported by tcpl. tcplfit2, however, can be used for stand-alone applications, available at https://cran.r-project.org/package=tcplfit2
Endpoint structure and annotation	Tcpl only fit in the positive analysis direction therefore dual endpoints were registered to capture gain and loss of signal	Given bidirectional fitting, a single endpoint is sufficient to capture both gain and loss of signal. Many endpoints were removed and/or renamed, and annotations were updated to reflect this paradigm shift. Continued curation efforts enable better data aggregation
Schema changes	Processed data was previously stored in “wide” format with a fixed number of columns in the level 4 (mc4) and level 5 (mc5) tables based on three curve-fitting models	Complete tcplfit2 model parameters are captured within the mc4_param and mc5_param tables, allowing for generic fitting and hit calling, with summary-level statistics now only stored in mc4 and mc5
Plotting	Several functions were used to produce the different plotting outputs	tcplPlot() allows for interactive, yet consistent visualization of concentration-response curves. As a new stand-alone plotting utility built with plotly to display the additional curve-fitting models, the utility implements plumber to provide REST API functionality, which can support Docker integration and web hosting

A summary description of major updates to invitrodb in v4.0 and tcpl in v3.0 are provided in Table 4, including updates to curve-fitting models, hitcall determination, potency estimation, data pipelining independent of invitrodb (“stand-alone pipelining”), endpoint structure and annotation, changes to the invitrodb schema, and new plotting.