R/fitMultiCurves.R

In BaileyLabUM/biosensor: Analyzes raw data from biosensors, specifically, the Maverick M1 detection system from Genalyte, Inc

fitMultiCurves <- function(data, loc){
        ggplot2::theme_set(ggthemes::theme_few(base_size = 16))

        tarList <- unique(data$Target)
        fit <- list()
        for(i in tarList) {
                tarDat <- dplyr::filter(data, Target == i)
                tarDat$Cycle <- as.numeric(tarDat$Cycle)
                startA <- max(tarDat$NetShift)
                startB <- min(tarDat$NetShift)
                startC <- median(tarDat$NetShift)
                startD <- 1
                y <- tarDat$NetShift
                x <- as.numeric(tarDat$Cycle)
                nls.control(minFactor = 5)
                # tarFit <- tryCatch({fit.info <- nls(formula = y ~ A + (B - A) /
                #                                             (1 + (x / C) ^ D),
                #                                     start = list(A = startA,
                #                                                  B = startB,
                #                                                  C = startC,
                #                                                  D = startD))
                # },

                tarFit <- tryCatch({fit.info <- glm(y ~ x)
                },
                error = function(e) {"failed"},
                # finally = print(paste(i, startA, startB, startC, startD)))
                finally = print(i))
                if(tarFit[1] != "failed"){
                        fit[[i]] <- broom::tidy(fit.info)
                        fit[[i]]$Target <- unique(tarDat$Target)
                        # A <- as.numeric(coef(fit.info)[1])
                        # B <- as.numeric(coef(fit.info)[2])
                        # C <- as.numeric(coef(fit.info)[3])
                        # D <- as.numeric(coef(fit.info)[4])

                        # testFun <- function(x) {A + (B - A) / (1 + (x / C) ^ D)}

                        testFun <- function(x)

                        plot <- ggplot2::ggplot(tarDat,
                                                ggplot2::aes(x = Cycle,
                                                             y = NetShift,
                                                             group = Cycle)) +
                                ggplot2::geom_boxplot(fill = "red") +
                                ggplot2::stat_function(fun = testFun,
                                                       color = "blue", size = 1) +
                                ggplot2::labs(x = "Cycle Number",
                                              y = expression(paste("Relative Shift (",
                                                                   Delta,
                                                                   "pm)")))

                        ggplot2::ggsave(plot,
                                        filename = paste0(i, "CalCurve.png"),
                                        width = 8, height = 6)
                }
        }

        fit <- dplyr::bind_rows(fit)

        capture.output(fit, file = "fitInfo.txt")
        readr::write_csv(fit, path = "fitInfo.csv")
}