Skip to main content

Table 3 MAE and MEDAE results weighted CatBoost regressors with large overall weights for non-retained molecules (1/2, 1/1 and 2/1)

From: Probabilistic metabolite annotation using retention time prediction and meta-learned projections

Features Metric CB 1/2 CB 1/1 CB 2/1
desc fgp desc+fgp desc fgp desc+fgp desc fgp desc+fgp
All MAE (s) \(57.78\pm 1.04\) \(58.37\pm 0.82\) \(55.34\pm 1.02\) \(59.02\pm 1.11\) \(59.52\pm 0.87\) \(56.67\pm 0.89\) \(60.90\pm 1.17\) \(61.03\pm 0.96\) \(58.49\pm 1.16\)
MEDAE (s) \(34.40\pm 0.44\) \(35.02\pm 0.54\) \(32.39\pm 0.69\) \(35.62\pm 0.74\) \(36.04\pm 0.61\) \(33.71\pm 0.61\) \(37.46\pm 0.66\) \(37.48\pm 0.75\) \(35.49\pm 0.96\)
Non-retained MAE (s) \(168.46\pm 8.31\) \(169.78\pm 7.59\) \(167.70\pm 8.14\) \(166.46\pm 8.14\) \(166.81\pm 7.80\) \(165.52\pm 7.85\) \(164.59\pm 7.98\) \(165.19\pm 7.51\) \(162.96\pm 7.96\)
MEDAE (s) \(16.12\pm 2.20\) \(15.99\pm 2.05\) \(15.34\pm 2.65\) \(12.32\pm 2.20\) \(10.93\pm 1.42\) \(11.48\pm 1.94\) \(9.59\pm 1.30\) \(9.10\pm 1.11\) \(9.49\pm 1.32\)
Retained MAE (s) \(54.86\pm 1.14\) \(55.43\pm 0.87\) \(52.38\pm 1.08\) \(56.19\pm 1.15\) \(56.69\pm 0.90\) \(53.80\pm 0.94\) \(58.16\pm 1.25\) \(58.28\pm 1.04\) \(55.74\pm 1.24\)
MEDAE (s) \(34.71\pm 0.41\) \(35.38\pm 0.53\) \(32.70\pm 0.63\) \(36.00\pm 0.65\) \(36.43\pm 0.57\) \(34.06\pm 0.57\) \(37.88\pm 0.65\) \(37.93\pm 0.72\) \(35.94\pm 0.91\)
  1. In this table, desc, fgp and desc+fgp refer to the input features used by each regressor. desc means descriptors, fgp means fingerprints and desc+fgp means that both descriptors and fingerprints have been used