Figure 3.

Summary of the data at each step of the workflow. Row 1: (Step 1) Inter-dataset distances for matched features in the (RT, MZ, log₁₀FI) domains. Black dots are unique matches, blue circles are matches in clusters, and orange dots are matches outside the log₁₀FI threshold limits. Row 2: (Step 2a) Black dots are the same as in Row 1, red circles are expected values at the (RT, MZ, log₁₀FI) of the reference feature in the match. Row 3: (Step 2b) Residuals of the expected values. Row 4: (additional Step 2b) Normalized residuals obtained by dividing by the threshold point at their median + 3 × MAD. Row 5: (Step 2d) After defining weights W = [1,1,0.2] (Step 2c, not shown) penalization scores are obtained and used to color the same plots as in Row 1 (RT and MZ) and the comparison of log₁₀FI of target and reference. Penalization scores are used (Step 2e, not shown) to decide the best match in clusters with multiple matches. Row 6: (Step 3) Tightening of thresholds used to define poor matches using the method “scores” at the threshold limit of median + 3 × MAD. Matches (part of clusters) previously discarded in blue, poor matches in red, and good matches in black.