Package 'tcplfit2'

Description

GNLS objective function set to y for gnls solver.

Usage

acgnlsobj(x, y, tp, ga, p, la, q)

Arguments

Value

Difference between GNLS model repsone at x and y.

Description

Returns concentration at which model equals y.

Usage

acy(
  y,
  modpars,
  type = "hill",
  returntop = FALSE,
  returntoploc = FALSE,
  getloss = FALSE,
  poly2.biphasic = TRUE,
  verbose = FALSE
)

Arguments

Details

Mathematically inverts model functions of the given type, except for gnls, which is numerically inverted. gnls returns NA when y > tp. Other options return the actual top (as opposed to theoretical tp) and top location for gnls model. gnls model defaults to giving concentration on gain side. Only one of getloss, returntop, and returntoploc should be TRUE at a time. If top location solution fails for gnls, top is set to tp. Returns NA if gnls numerical solver fails. Returns NA if model was not successfully fit.

Value

Ouputs concentration at activity y, or gnls top or top concentration, when applicable.

Examples

acy(1, list(ga = 10, tp = 2, p = 3), type = "hill")
acy(1, list(ga = .1, tp = 2, p = 3, q = 3,la = 10), type = "gnls")
acy(1, list(ga = .1, tp = 2, p = 3, q = 3,la = 10), type = "gnls", getloss = TRUE)
acy(1, list(ga = .1, tp = 2, p = 3, q = 3,la = 10), type = "gnls", returntop = TRUE)
acy(1, list(ga = .1, tp = 2, p = 3, q = 3,la = 10), type = "gnls", returntoploc = TRUE)

Description

Uses maximum likelihood method to tune the upper and lower bounds on the BMD (BMDU, BMDL)

Usage

bmdbounds(
  fit_method,
  bmr,
  pars,
  conc,
  resp,
  onesidedp = 0.05,
  bmd = NULL,
  which.bound = "lower",
  poly2.biphasic = TRUE,
  x_v
)

Arguments

Details

Takes in concentration response fit details and outputs a bmdu or bmdl, as desired. If bmd is not finite, returns NA. If the objective function doesn't change sign or the root finding otherwise fails, it returns NA. These failures are not uncommon since some curves just don't reach the desired confidence level.

Value

Returns either the BMDU or BMDL.

Examples

conc = c(.03, .1, .3, 1, 3, 10, 30, 100)
resp = c(.1,-.1,0,1.1,1.9,2,2.1,1.9)
pars = c(tp = 1.973356, ga = 0.9401224, p = 3.589397, er =  -2.698579)
bmdbounds(fit_method = "hill", bmr = .5, pars, conc, resp)
bmdbounds(fit_method = "hill", bmr = .5, pars, conc, resp, which.bound = "upper")

Description

Utility function for bmdbounds

Usage

bmdobj(
  bmd,
  fname,
  bmr,
  conc,
  resp,
  ps,
  mll,
  onesp,
  partype = 2,
  poly2.biphasic = TRUE,
  x_v
)

Arguments

Value

Objective function value to find the zero of.

Description

$f(x) = 0$

Usage

cnst(ps, x)

Arguments

Value

Vector of model responses

Examples

cnst(1,1)

Description

Core of concentration response curve fitting for pvalue based cutoff. This function calls tcplfit2_core to get curve fits, and then tcplhit2_core to perform the hitcalling. Prior to model fitting, this function includes two data preparation steps (1) centering responses when bmed is not 0 or NULL and (2) removal of replicates with missing response values.

Usage

concRespCore(
  row,
  fitmodels = c("cnst", "hill", "gnls", "poly1", "poly2", "pow", "exp2", "exp3", "exp4",
    "exp5"),
  conthits = TRUE,
  aicc = FALSE,
  force.fit = FALSE,
  bidirectional = TRUE,
  verbose = FALSE,
  do.plot = FALSE,
  return.details = FALSE,
  errfun = "dt4",
  bmr_scale = 1.349,
  bmd_low_bnd = NULL,
  bmd_up_bnd = NULL,
  poly2.biphasic = TRUE,
  AUC = FALSE,
  use.abs.auc = FALSE,
  use.log.auc = FALSE
)

Arguments

Value

A list of two elements. The first (summary) is the output from tcplhit2_core. The second, params is the output from tcplfit2_core a dataframe of one row containing

Examples

conc <- list(.03, .1, .3, 1, 3, 10, 30, 100)
resp <- list(0, .2, .1, .4, .7, .9, .6, 1.2)
row <- list(conc = conc,
            resp = resp,
            bmed = 0,
            cutoff = 1,
            onesd = .5,
            name = "some chemical",
            assay = "some assay")
concRespCore(row, conthits = TRUE)
concRespCore(row, aicc = TRUE)

Description

Plots a concentration response curve for one sample/endpoint combination. This is a generic function and it is expected that users will make their own versions

Usage

concRespPlot(row, ymin = -120, ymax = 120, draw.error.arrows = FALSE)

Arguments

Details

row is one row of data from concRespCore

Value

No output.

Examples

conc <- list(.03, .1, .3, 1, 3, 10, 30, 100)
resp <- list(0, .2, .1, .4, .7, .9, .6, 1.2)
row <- list(conc = conc,
            resp = resp,
            bmed = 0,
            cutoff = 0.25,
            onesd = 0.125,
            name = "some chemical",
            assay = "some assay")
res <- concRespCore(row, conthits = TRUE)
concRespPlot(res,ymin=-2.5,ymax=2.,5)

Description

This function takes output from 'concRespCore' or 'tcplhit2_core' to generate a basic plot of the observed concentration-response data and the best fit curve (winning model). A 'ggplot' object, which users may customize with additional 'ggplot' layers, is returned.

Usage

concRespPlot2(row, log_conc = FALSE)

Arguments

Value

A 'ggplot' object of the observed concentration-response data overlaid with the best fit curve (winning model).

Description

$f(x) = a*(e^{(x/b)}- 1)$

Usage

exp2(ps, x)

Arguments

Value

Vector of model responses

Examples

exp2(c(1,2),1)

Description

$f(x) = a*(e^{(x/b)^p} - 1)$

Usage

exp3(ps, x)

Arguments

Value

Vector of model responses

Examples

exp3(c(1,2,2),1)

Description

$f(x) = tp*(1-2^{(-x/ga)})$

Usage

exp4(ps, x)

Arguments

Value

Vector of model responses

Examples

exp4(c(1,2),1)

Description

$f(x) = tp*(1-2^{(-(x/ga)^p)})$

Usage

exp5(ps, x)

Arguments

Value

Vector of model responses

Examples

exp5(c(1,2,3),1)

Description

Function that fits a constant line $f(x) = 0$ and returns generic model outputs.

Usage

fitcnst(conc, resp, nofit = FALSE, errfun = "dt4", ...)

Arguments

Details

success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. aic, rme, and er are set to NA in case of nofit or failure. pars always equals "er".

Value

List of five elements: success, aic (Akaike Information Criteria), rme (root mean square error), er (error parameter), pars (parameter names).

Examples

fitcnst(c(.1,1,10,100), c(1,2,0,-1))
fitcnst(c(.1,1,10,100), c(1,2,0,-1), nofit = TRUE)

Description

Function that fits to $f(x) = a*(e^{(x/b)}- 1)$ and returns generic model outputs.

Usage

fitexp2(
  conc,
  resp,
  bidirectional = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  errfun = "dt4"
)

Arguments

Details

Zero background and increasing absolute response are assumed. Parameters are "a" (y scale), "b" (x scale), and error term "er". success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fitexp2(c(.1,1,10,100), c(0,.1,1,10))

Description

Function that fits to $f(x) = a*(e^{(x/b)^p} - 1)$ and returns generic model outputs.

Usage

fitexp3(
  conc,
  resp,
  bidirectional = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  dmin = 0.3,
  errfun = "dt4"
)

Arguments

Details

Zero background and increasing absolute response are assumed. Parameters are "a" (y scale), "b" (x scale), "p" (power), and error term "er". success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fitexp3(c(.03,.1,.3,1,3,10,30,100), c(0,0,.1, .2, .4, 1, 4, 50))

Description

Function that fits to $f(x) = tp*(1-2^{(-x/ga)})$ and returns generic model outputs.

Usage

fitexp4(
  conc,
  resp,
  bidirectional = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  errfun = "dt4"
)

Arguments

Details

Zero background and increasing absolute response are assumed. Parameters are "tp" (top), "ga" (AC50), and error term "er". success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fitexp4(c(.03,.1,.3,1,3,10,30,100), c(0,0,.1, .2, .5, 1, 1.5, 2))

Description

Function that fits to $f(x) = tp*(1-2^{(-(x/ga)^p)})$ and returns generic model outputs.

Usage

fitexp5(
  conc,
  resp,
  bidirectional = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  dmin = 0.3,
  errfun = "dt4"
)

Arguments

Details

Zero background and increasing absolute response are assumed. Parameters are "tp" (top), "ga" (AC50), "p" (power), and error term "er". success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fitexp5(c(.03,.1,.3,1,3,10,30,100), c(0,0,.1, .2, .5, 1, 1.5, 2))

Description

Function that fits to $f(x) = \frac{tp}{[(1 + (ga/x)^p )(1 + (x/la)^q )]}$ and returns generic model outputs.

Usage

fitgnls(
  conc,
  resp,
  bidirectional = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  minwidth = 1.5,
  errfun = "dt4"
)

Arguments

Details

Concentrations are converted internally to log10 units and optimized with $f(x) = \frac{tp}{[(1 + 10^{(p*(ga-x))} )(1 + 10^{(q*(x-la))} )]}$, then ga, la, ga_sd, and la_sd are converted back to regular units before returning. Zero background and increasing initial absolute response are assumed. Parameters are "tp" (top), "ga" (gain AC50), "p" (gain power), "la" (loss AC50),"q" (loss power) and error term "er". success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fitgnls(c(.03,.1,.3,1,3,10,30,100), c(0,.3,1, 2, 2.1, 1.5, .8, .2))

Description

Function that fits to $f(x) = \frac{tp}{[(1 + (ga/x)^p )]}$ and returns generic model outputs.

Usage

fithill(
  conc,
  resp,
  bidirectional = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  errfun = "dt4"
)

Arguments

Details

Concentrations are converted internally to log10 units and optimized with $f(x) = \frac{tp}{(1 + 10^{(p*(ga-x))} )}$, then ga and ga_sd are converted back to regular units before returning. Zero background and increasing initial absolute response are assumed. Parameters are "tp" (top), "ga" (gain AC50), "p" (gain power), and error term "er". success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fithill(c(.03,.1,.3,1,3,10,30,100), c(0,0,.1, .2, .5, 1, 1.5, 2))

Description

Function that fits to $f(x) = a*x$ and returns generic model outputs.

Usage

fitpoly1(
  conc,
  resp,
  bidirectional = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  errfun = "dt4"
)

Arguments

Details

Zero background and increasing absolute response are assumed. Parameters are "a" (y scale) and error term "er". success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fitpoly1(c(.03,.1,.3,1,3,10,30,100), c(0,.01,.1, .1, .2, .5, 2, 5))

Description

Function that fits to $f(x) = b1*x + b2*x^2$ (biphasic), or $f(x) = a*(\frac{x}{b} + \frac{x^2}{b^2})$ (monotonic only), and returns generic model outputs.

Usage

fitpoly2(
  conc,
  resp,
  bidirectional = TRUE,
  biphasic = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  errfun = "dt4"
)

Arguments

Details

(Biphasic Poly2 Model) Zero background is assumed and responses may be biphasic (non-monotonic). Parameters are "b1" (shift along x-axis), "b2" (rate of change, direction, and the shift along y-axis), and error term "er". (Monotonic Poly2 Model) Zero background and monotonically increasing absolute response are assumed. Parameters are "a" (y scale), "b" (x scale), and error term "er". (Biphasic or Monotonic Poly2 Fit) success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fitpoly2(c(.03,.1,.3,1,3,10,30,100), c(0,.01,.1, .1, .2, .5, 2, 8))

Description

Function that fits to $f(x) = a*x^p$ and returns generic model outputs.

Usage

fitpow(
  conc,
  resp,
  bidirectional = TRUE,
  verbose = FALSE,
  nofit = FALSE,
  nmin = 0.3,
  errfun = "dt4"
)

Arguments

Details

Zero background and monotonically increasing absolute response are assumed. Parameters are "a" (y scale), "p" (power), and error term "er". success = 1 for a successful fit, 0 if optimization failed, and NA if nofit = TRUE. cov = 1 for a successful hessian inversion, 0 if it fails, and NA if nofit = TRUE. aic, rme, modl, parameters, and parameter sds are set to NA in case of nofit or failure.

Value

Named list containing: success, aic (Akaike Information Criteria), cov (success of covariance calculation), rme (root mean square error), modl (vector of model values at given concentrations), parameters values, parameter sd (standard deviation) estimates, pars (vector of parameter names), sds (vector of parameter sd names).

Examples

fitpow(c(.03,.1,.3,1,3,10,30,100), c(0,.01,.1, .1, .2, .5, 2, 8))

Description

Function that calculates the area under the curve (AUC) for dose-response curves.

Usage

get_AUC(fit_method, lower, upper, ps, return.abs = FALSE, use.log = FALSE)

Arguments

Details

This function takes in a model name and the respective set of model parameters, and returns the area under the curve (AUC) between the specified lower and upper concentration bounds. The AUC can be used to compute an efficacy/potency metric for "active" dose-response curves. For decreasing curves, the AUC returned will be negative. However, users have the option to return a positive AUC in these cases. Model parameters should be entered as a numeric list or vector. Models optimized on the log10-scale (hill and gain-loss), the lower and upper concentration bounds, parameters "ga" (gain AC50) and "la" (loss AC50) will be converted to log10-scale.

Value

AUC value (numeric)

Examples

conc <- c(.03,.1,.3,1,3,10,30,100)
fit_method <- "gnls"
modpars <- list(tp = 1.023, ga = 2.453, p = 1.592,
                la = 4288.993, q = 5.770, er = -3.295)
get_AUC("exp2", min(conc), max(conc), ps = modpars)

Description

$f(x) = \frac{tp}{[(1 + (ga/x)^p )(1 + (x/la)^q )]}$

Usage

gnls(ps, x)

Arguments

Value

Vector of model responses

Examples

gnls(c(1,2,1,2,2),1)

Description

Derivative of the gnls function set to zero for top location solver.

Usage

gnlsderivobj(x, tp, ga, p, la, q)

Arguments

Value

Value of gnls derivative at x.

Description

$f(x) = \frac{tp}{[(1 + (ga/x)^p )]}$

Usage

hillfn(ps, x)

Arguments

Value

Vector of model responses

Examples

hillfn(c(1,2,3),1)

Description

Wrapper that computes continuous hitcalls for a provided concRespCore input row.

Usage

hitcont(indf, xs = NULL, ys = NULL, newcutoff)

Arguments

Details

indf parameter columns should be NA when not required by fit method. "conc" and "resp" entries should be a single string with values separated by |. Details on indf columns can be found in concRespCore.

Value

Vector of hitcalls between 0 and 1 with length equal to indf row number.

Examples

conc <- list(.03, .1, .3, 1, 3, 10, 30, 100)
resp <- list(0, .2, .1, .4, .7, .9, .6, 1.2)
row <- list(
  conc = conc,
  resp = resp,
  bmed = 0,
  cutoff = 1,
  onesd = .5,
  name = "some chemical",
  assay = "some assay"
)
res <- concRespCore(row, conthits = TRUE)
hitcont(res, newcutoff = 0.2)

Description

Calculates continuous hitcall using 3 statistical metrics.

Usage

hitcontinner(
  conc,
  resp,
  top,
  cutoff,
  er,
  ps,
  fit_method,
  caikwt,
  mll,
  errfun = "dt4"
)

Arguments

Details

This function is called either directly from concRespCore or via hitcont. Details of how to compute function input are in concRespCore.

Value

Continuous hitcall between 0 and 1.

Examples

conc = c(.03,.1,.3,1,3,10,30,100)
resp = c(0,.1,0,.2,.6,.9,1.1,1)
top = 1.023239
er = -3.295307
ps = c(1.033239, 2.453014, 1.592714, er = -3.295307) #tp,ga,p,er
fit_method = "hill"
caikwt = 1.446966e-08
mll = 12.71495
hitcontinner(conc,resp,top,cutoff = 0.8, er,ps,fit_method, caikwt, mll)
hitcontinner(conc,resp,top,cutoff = 1, er,ps,fit_method, caikwt, mll)
hitcontinner(conc,resp,top,cutoff = 1.2, er,ps,fit_method, caikwt, mll)

Description

Wrapper that computes discrete hitcalls for a provided concRespCore dataframe.

Usage

hitlogic(indf, newbmad = NULL, xs = NULL, ys = NULL, newcutoff = NULL)

Arguments

Value

Vector of hitcalls with length equal to number of rows in indf.

Examples

conc = rep(".03|.1|.3|1|3|10|30|100",2)
resp = rep("0|0|.1|.1|.5|.5|1|1",2)
indf = data.frame(top = c(1,1), ac50 = c(3,4), conc = conc, resp = resp,
  stringsAsFactors = FALSE)
hitlogic(indf, newcutoff = c(.8,1.2))

Description

Contains hit logic, called directly during CR fitting or later through "hitlogic".

Usage

hitloginner(conc = NULL, resp, top, cutoff, ac50 = NULL)

Arguments

Details

The purpose of this function is to keep the actual hit rules in one location so it can be called during CR fitting, and then again after the fact for a variety of cutoffs. Curves fit with constant winning should have top = NA, generating a miss.

Value

Outputs 1 for hit, 0 for miss.

Examples

hitloginner(resp = 1:8, top = 7, cutoff = 5) #hit
hitloginner(resp = 1:8, top = 7, cutoff = 7.5) #miss: top too low
hitloginner(resp = 1:8, top = 9, cutoff = 8.5) #miss: no response> cutoff
hitloginner(resp = 1:8, top = NA, cutoff = 5) #miss: no top (constant)

Description

$f(x) = \frac{tp}{[(1 + 10^{(p*(ga-x))} )(1 + 10^{(q*(x-la))} )]}$

Usage

loggnls(ps, x)

Arguments

Value

Vector of model responses

Examples

loggnls(c(1,2,1,2,2),1)

Description

$f(x) = \frac{tp}{(1 + 10^{(p*(ga-x))} )}$

Usage

loghill(ps, x)

Arguments

Value

Vector of model responses.

Examples

loghill(c(1,2,3),1)

Description

A data set containing 1831 chemicals worth of data for the ACEA_AR assay, Data from the assay component ACEA_AR_agonist_80hr was analyzed in the positive analysis fitting direction relative to DMSO as the neutral control and baseline of activity. The data can be accessed further through invitrodb and tcpl see https://www.epa.gov/chemical-research/exploring-toxcast-data#Download.

Usage

mc0

Format

An object of class data.table (inherits from data.frame) with 53608 rows and 13 columns.

Details

This data is extracted from the released version of the ToxCast database, invitrodb, at level 0 (mc0) and contains the concentration-response information.

A data frame with 53608 rows and 13 variables:

• m0id - Level 0 id

• spid - Sample id

• acid - Unique assay component id; unique numeric id for each assay component

• apid - Assay plate id

• rowi - Row index (location on assay plate)

• coli - Column index (location on assay plate)

• wllt - Well type

• wllq - well quality

• conc - concentration

• rval - raw value

• srcf - Source file name

• clowder_uid - clowder unique id for source files

• git_hash - hash key for pre-processing scripts

Source

doi:10.23645/epacomptox.6062623.v10

Description

A data set containing 100 chemicals worth of data for the Tox21 assay TOX21_ERa_BLA_Agonist_ratio, which measures response to estrogen receptor agonists. The data can be accessed further through the Comptox Chemicals Dashboard (https://comptox.epa.gov/dashboard).

Usage

mc3

Format

An object of class data.frame with 32175 rows and 7 columns.

Details

This data is extracted from the released version of the ToxCast database, invitrodb, at level 3 (mc3) and contains the concentration-response information.

A data frame with 32175 rows and 7 variables:

• dtxsid - DSSTox generic substance ID

• casrn - Chemical Abstracts Registry Number (CASRN)

• name - chemical name

• spid - sample ID - there can be multiple samples per chemical

• logc - log10(concentration), micromolar (uM)

• resp - response in %

• assay - name of the assay / assay component endpoint name

Source

doi:10.23645/epacomptox.6062623.v5

Description

Chooses between nested models.

Usage

nestselect(aics, mod1, mod2, dfdiff, pval = 0.05)

Arguments

Value

Named aic vector with losing model removed.

Examples

aics = c(-5,-6,-3)
names(aics) = c("poly1", "poly2", "hill")
nestselect(aics, "poly1", "poly2", 1)

aics = c(-5,-7,-3)
names(aics) = c("poly1", "poly2", "hill")
nestselect(aics, "poly1", "poly2", 1)

Description

This function takes output from 'tcplfit2_core' and generates a basic plot of the observed concentration-response data with all resulting curve fits. A 'ggplot' object, which users may customize with additional 'ggplot' layers, is returned.

Usage

plot_allcurves(modelfits, conc, resp, log_conc = FALSE)

Arguments

Value

A 'ggplot' object of the observed concentration-response data and all resulting curve fits from 'tcplfit2_core'. (Note: The constant model is not included, and only the successful fits will be displayed.)

Description

$f(x) = a*x$

Usage

poly1(ps, x)

Arguments

Value

Vector of model responses

Examples

poly1(1,1)

Description

$f(x) = a*(\frac{x}{b} + \frac{x^2}{b^2})$

Usage

poly2(ps, x)

Arguments

Value

Vector of model responses

Examples

poly2(c(1,2),1)

Description

$f(x) = b1*x + b2*x^{2}$

Usage

poly2bmds(ps, x)

Arguments

Value

Vector of model responses

Examples

poly2bmds(c(1,2),1)

Description

Function that calculates the area under the curve (AUC) after hit-calling.

Usage

post_hit_AUC(hit_results, return.abs = FALSE, use.log = FALSE)

Arguments

Details

This function calculates the area under the curve (AUC) for the winning model selected during hit-calling. Wrapper function for 'get_AUC'. Designed to take the one-row output from 'tcplhit2_core', parse the model details, and pass these values to 'get_AUC' to estimate the AUC for the winning model.

Value

AUC value of the winning model (numeric)

See Also

get_AUC

Examples

conc <- c(.03, .1, .3, 1, 3, 10, 30, 100)
resp <- c(0, .2, .1, .4, .7, .9, .6, 1.2)
params <- tcplfit2_core(conc, resp, .8)
output <- tcplhit2_core(params, conc, resp, 0.8, 0.5)
post_hit_AUC(output)

Description

$f(x) = a*x^p$

Usage

pow(ps, x)

Arguments

Value

Vector of model responses

Examples

pow(c(1,2),1)

Description

A data set containing 6 of the active transcriptional signatures after perturbation of MCF7 cells with Clomiphene citrate (1:1).

Usage

signatures

Format

An object of class data.frame with 6 rows and 8 columns.

Details

A data frame with 6 rows and 8 variables:

• sample_id - experimental sample ID

• dtxsid - DSSTox generic substance ID

• name - chemical name

• signature - transcriptional signature name

• cutoff - the 95% confidence interval from the baseline response (2 lowest concentrations)

• onesd - one standard deviation of the baseline response

• conc - experimental concentrations, micromolar (uM)

• resp - transcriptional signature response for each experimental concentrations, ssGSEA score

Source

doi:10.1093/toxsci/kfab009

References

Joshua A. Harrill, Logan J. Everett, Derik E. Haggard, Thomas Sheffield, Joseph L. Bundy, Clinton M. Willis, Russell S. Thomas, Imran Shah, Richard S. Judson, High-Throughput Transcriptomics Platform for Screening Environmental Chemicals, Toxicological Sciences, Volume 181, Issue 1, May 2021, Pages 68 - 89, https://doi.org/10.1093/toxsci/kfab009.

Description

Concentration response curve fitting using the methods from BMDExpress

Usage

tcplfit2_core(
  conc,
  resp,
  cutoff,
  force.fit = FALSE,
  bidirectional = TRUE,
  verbose = FALSE,
  do.plot = FALSE,
  fitmodels = c("cnst", "hill", "gnls", "poly1", "poly2", "pow", "exp2", "exp3", "exp4",
    "exp5"),
  poly2.biphasic = TRUE,
  errfun = "dt4",
  ...
)

Arguments

Details

All models are equal to 0 at 0 concentration (zero background). To add more models in the future, write a fit____ function, and add the model name to the fitmodels and modelnames vectors.

Value

List of N(models) elements, one for each of the models run (up to 10), followed by a element "modelnames", which is a vector of model names so other functions can easily cycle through the output, and then the last element "errfun", which indicates what distribution was used for error. For a full list, see the documentation for the individual fitting method functions. For each model there is a sublist with elements including:

• success - was the model successfully fit

• aic - the AIC value

• cov - success of the the covariance matrix calculation

• rme - root mean error of the data around the curve

• modl - vector of model values at the given concentrations

• tp - the top of the curve fit

• ga - the AC50 or Hill paramters

• er - the error term

• ... other paramters specific to the model (see the documentation for the specific models)

• tp_sd, ga_sd, p_sd, etc., the values of the standard deviations of the paramters for the models

• er_sd - standard deviation of the error term

• pars - the names of the parameters

• sds - the names of the standard deviations of the paramters

Examples

conc <- c(.03, .1, .3, 1, 3, 10, 30, 100)
resp <- c(0, .1, 0, .2, .6, .9, 1.1, 1)
output <- tcplfit2_core(conc, resp, .8,
  fitmodels = c("cnst", "hill"), verbose = TRUE,
  do.plot = TRUE
)

Description

Core of hitcalling function. This method chooses the winning model from tcplfit2_core, extracts the top and ac50, computes the hitcall, and calculates bmd/bmdl/bmdu among other statistics. Nested model selection is used to choose between poly1/poly2, then the model with the lowest AIC (or AICc) is declared the winner. Continuous hitcalls requires tcplfit2_core to be run with force.fit = TRUE and "cnst" never to be chosen as the winner.

Usage

tcplhit2_core(
  params,
  conc,
  resp,
  cutoff,
  onesd,
  bmr_scale = 1.349,
  bmed = 0,
  conthits = TRUE,
  aicc = FALSE,
  identifiers = NULL,
  bmd_low_bnd = NULL,
  bmd_up_bnd = NULL,
  poly2.biphasic = TRUE
)

Arguments

Value

A list of with the detailed results from all of the different model fits. The elements of summary are:

• any elements of the identifiers input

• n_gt_cutoff - number of data points above the cutoff

• cutoff - noise cutoff

• fit_method - curve fit method

• top_over_cutoff - top divided by cutoff

• rmse - RMSE of the data points around the best model curve

• a - fitting parameter methods: exp2, exp3, poly1, poly2, pow

• b - fitting parameter methods: exp2, exp3, ploy2

• p - fitting parameter methods: exp3, exp5, gnls, hill, pow

• q - fitting parameter methods: gnls,

• tp - top of the curve

• ga - ac50 for the rising curve in a gnls model or the Hill model

• la - ac50 for the falling curve in a gnls model

• er - fitted error term for plotting error bars

• bmr - benchmark response; level at which bmd is calculated = onesd*bmr_scale default bmr_scale is 1.349

• bmd - benchmark dose, curve value at bmr

• bmdl - lower limit on the bmd

• bmdu - upper limit on the bmd

• caikwt - one factor used in calculating the continuous hitcall. It is calcalated from the formula = exp(-aic(cnst)/2)/(exp(-aic(cnst)/2) + exp(-aic(fit_method)/2)) and measures how much lower the selected method AIC is than that for the constant model

• mll - another factor used in calcualting the continuous hitcall = length(modpars) - aic(fit_method)/2

• hitcall - the final hitcall, a value ranging from 0 to 1

• top - curve top

• ac50 - curve value at 50% of top, curve value at cutoff

• lc50 - curve value at 50% of top corresponding to the loss side of the gain-loss curve

• ac5 - curve value at 5% of top

• ac10 - curve value at 10% of top

• ac20 - curve value at 20% of top

• acc - curve value at cutoff

• ac1sd - curve value at 1 standard deviation

• conc - conc string separated by |'s

• resp - response string separated by |'s

Description

Log-likelihood to be maximized during CR fitting.

Usage

tcplObj(p, conc, resp, fname, errfun = "dt4", err = NULL)

Arguments

Details

This function is a generalized version of the log-likelihood estimation functions used in the ToxCast Pipeline (TCPL). Hill model uses fname "loghill" and gnls uses fname "loggnls". Other model functions have the same fname as their model name; i.e. exp2 uses "exp2", etc. errfun = "dnorm" may be better suited to gsva pathway scores than "dt4". Setting err could be used to fix error based on the null data noise distribution instead of fitting the error when maximizing log-likelihood.

Value

Log-likelihood.

Examples

conc = c(.03,.1 , .3 , 1  , 3 , 10 , 30  , 100)
resp = c( 0 , 0 , .1 ,.2 , .5 , 1  , 1.5 , 2  )
p = c(tp = 2, ga = 3, p = 4, er = .5)
tcplObj(p,conc,resp,"exp5")

lconc = log10(conc)
tcplObj(p,lconc,resp,"loghill")

Description

Probability of top being above cutoff.

Usage

toplikelihood(fname, cutoff, conc, resp, ps, top, mll, errfun = "dt4")

Arguments

Details

Should only be called by hitcontinner. Uses profile likelihood, similar to bmdbounds. Here, the y-scale type parameter is substituted in such a way that the top equals the cutoff. Then the log-likelihood is compared to the maximum log-likelihood using chisq function to retrieve probability.

Value

Probability of top being above cutoff.

Examples

fname = "hillfn"
conc = c(.03,.1,.3,1,3,10,30,100)
resp = c(0,.1,0,.2,.6,.9,1.1,1)
ps = c(1.033239, 2.453014, 1.592714, er = -3.295307)
top = 1.023239
mll = 12.71495
toplikelihood(fname, cutoff = .8, conc, resp, ps, top, mll)
toplikelihood(fname, cutoff = 1, conc, resp, ps, top, mll)
toplikelihood(fname, cutoff = 1.2, conc, resp, ps, top, mll)