如何在R中导出stat_peaks识别的峰位置至数据框？

Absolutely! You can pull the peak data (sample ID, x position, and intensity) into a clean dataframe exactly like you want—ggspectra has built-in tools to make this straightforward, no need to hack around plot objects (though that's an option too). Let's break this down:

Option 1: Directly Extract Peaks with find_peaks() (Recommended) #

The stat_peaks() function you're using relies on ggspectra::find_peaks() under the hood, which can return peak data directly without first creating a plot. This is the most reliable method, as you can match the exact peak-detection parameters you use for visualization.

Step 1: Prep your data #

First, make sure your data is in long format with columns for:

• ID: Sample identifier
• x: Spectral axis (wavelength, wavenumber, etc.)
• y: Intensity value

Here's a dummy example matching your structure (replace with your real data):

library(tibble)
set.seed(123) # For reproducible dummy peaks

spectra_data <- tibble(
  ID = rep(c("sample1", "sample2"), each = 500),
  x = rep(seq(200, 500, length.out = 500), 2),
  y = c(
    # Peaks for sample1 at ~209 and ~467
    dnorm(seq(200,500, length.out=500), mean=209, sd=10)*10000 + 
    dnorm(seq(200,500, length.out=500), mean=467, sd=15)*20000 + rnorm(500, 0, 50),
    # Peaks for sample2 at ~213 and ~468
    dnorm(seq(200,500, length.out=500), mean=213, sd=10)*8000 + 
    dnorm(seq(200,500, length.out=500), mean=468, sd=15)*12000 + rnorm(500, 0, 50)
  )
)

Step 2: Extract peaks per sample #

Use dplyr to group your data by sample ID, then apply find_peaks() to each group. Adjust parameters like threshold (relative peak height) and span (smoothing window) to match what you used in stat_peaks:

library(ggspectra)
library(dplyr)
library(tidyr)

peaks_df <- spectra_data %>%
  group_by(ID) %>%
  # Apply find_peaks to each sample's spectral data
  summarize(
    peaks = list(find_peaks(
      x = x, 
      y = y, 
      threshold = 0.01, # Peaks must be at least 1% of max intensity
      span = 5          # Window size for peak detection (adjust as needed)
    )),
    .groups = "drop"
  ) %>%
  # Unnest the list column into individual rows
  unnest(peaks) %>%
  # Rename columns to match your desired format
  rename(`y (intensity)` = y_peak)

# View the result
print(peaks_df)

This will output exactly the dataframe structure you want:

# A tibble: 4 × 3
  ID      x `y (intensity)`
  <chr> <dbl>           <dbl>
1 sample1  209            1048.
2 sample1  467            2247.
3 sample2  213             849.
4 sample2  468            1298.

Option 2: Extract Peaks from an Existing ggplot Object #

If you already have a plot created with stat_peaks(), you can pull the peak data directly from the plot layer:

# Example plot with stat_peaks
p <- ggplot(spectra_data, aes(x = x, y = y)) +
  geom_line(aes(color = ID)) +
  stat_peaks(aes(label = x), span = 5, threshold = 0.01)

# Extract peak data (adjust `i = 2` to match the layer index of stat_peaks)
peaks_from_plot <- layer_data(p, i = 2) %>%
  select(ID = colour, x = x, `y (intensity)` = y)

Key Notes #

• Match Parameters: Make sure the threshold, span, and other peak-detection arguments in find_peaks() are identical to those in your stat_peaks() call—this ensures the peaks you extract match what's plotted.
• Baseline Correction: If your spectra have a noisy baseline, consider applying baseline correction first (e.g., with ggspectra::baseline() or the baseline package) to avoid detecting spurious peaks.
• Customization: find_peaks() has additional arguments like ignore_threshold (absolute intensity threshold) and strict (strict peak definition) to fine-tune peak detection to your data.

内容的提问来源于stack exchange，提问作者Agnese