prof.to.features: parallelize central for loop

[hechth]

This loop could be paralelized to speed up processing. The number of processor should be passed as a parameter.

recetox-aplcms/R/prof.to.features.R

Lines 927 to 966 in a393e64

	# loop over each group
	for (i in seq_along(feature_groups))
	{
	# init variables
	feature_group <- feature_groups[[i]] |> dplyr::arrange_at("rt")
	num_features <- nrow(feature_group)
	# The estimation procedure for a single peak
	# Defines the dataframe containing median_mz, median_rt, sd1, sd2, and area
	if (num_features < 2) {
	time_weights <- all_diff_mean_rts[which(base.curve[, "base.curve"] %in% feature_group[2])]
	rt_peak_shape <- c(feature_group[1], feature_group[2], NA, NA, feature_group[3] * time_weights)
	peak_parameters <- rbind(peak_parameters, rt_peak_shape)
	} else {
	# find bandwidth for these particular range
	rt_range <- range(feature_group[, "rt"])
	bw <- min(max(bandwidth * (max(rt_range) - min(rt_range)), min_bandwidth), max_bandwidth)
	bw <- seq(bw, 2 * bw, length.out = 3)
	if (bw[1] > 1.5 * min_bandwidth) {
	bw <- c(max(min_bandwidth, bw[1] / 2), bw)
	}
	rt_profile <- compute_chromatographic_profile(feature_group, base.curve)
	if (shape_model == "Gaussian") {
	rt_peak_shape <- compute_gaussian_peak_shape(rt_profile, bw, component_eliminate, BIC_factor, aver_diff)
	} else {
	rt_peak_shape <- bigauss.mix(rt_profile, sigma_ratio_lim = sigma_ratio_lim, bw = bw, moment_power = moment_power, peak_estim_method = peak_estim_method, eliminate = component_eliminate, BIC_factor = BIC_factor)$param[, c(1, 2, 3, 5)]
	}
	if (is.null(nrow(rt_peak_shape))) {
	peak_parameters <- rbind(peak_parameters, c(median(feature_group[, "mz"]), rt_peak_shape))
	} else {
	for (m in 1:nrow(rt_peak_shape))
	{
	rt_diff <- abs(feature_group[, "rt"] - rt_peak_shape[m, 1])
	peak_parameters <- rbind(peak_parameters, c(mean(feature_group[which(rt_diff == min(rt_diff)), 1]), rt_peak_shape[m, ]))
	}
	}
	}
	}