Skip to main content

Table 2 MAE and MEDAE results for LightGBM and weighted CatBoost regressors with small overall weights for non-retained molecules (0, 1/80 and 1/40)

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

   LGB CB 0 CB 1/80 CB 1/40
Features Metric desc fgp desc+fgp desc fgp desc+fgp desc fgp desc+fgp desc fgp desc+fgp
All MAE (s) \(55.88\pm 1.03\) \(58.51\pm 0.42\) \(53.04\pm 1.01\) \(58.33\pm 0.52\) \(59.05\pm 0.54\) \(55.62\pm 0.44\) \(54.76\pm 0.62\) \(55.64\pm 0.44\) \(52.16\pm 0.59\) \(54.99\pm 0.70\) \(55.83\pm 0.49\) \(52.42\pm 0.67\)
MEDAE (s) \(32.92\pm 0.86\) \(35.61\pm 0.77\) \(30.60\pm 0.86\) \(32.24\pm 0.27\) \(32.95\pm 0.25\) \(29.99\pm 0.20\) \(32.34\pm 0.38\) \(33.21\pm 0.43\) \(30.04\pm 0.32\) \(32.39\pm 0.10\) \(33.42\pm 0.30\) \(30.25\pm 0.38\)
Non-retained MAE (s) \(220.95\pm 3.37\) \(221.34\pm 4.12\) \(219.66\pm 2.68\) \(487.59\pm 8.34\) \(486.92\pm 9.00\) \(484.03\pm 8.39\) \(238.73\pm 5.13\) \(241.84\pm 5.38\) \(237.98\pm 5.68\) \(215.67\pm 3.25\) \(217.78\pm 2.70\) \(214.64\pm 2.69\)
MEDAE (s) \(180.65\pm 15.02\) \(179.68\pm 38.10\) \(178.74\pm 24.13\) \(481.10\pm 7.02\) \(479.32\pm 8.16\) \(476.80\pm 8.67\) \(242.39\pm 16.82\) \(242.41\pm 24.63\) \(237.62\pm 17.12\) \(163.75\pm 14.74\) \(156.45\pm 28.25\) \(160.99\pm 17.41\)
Retained MAE (s) \(51.53\pm 1.11\) \(54.22\pm 0.44\) \(48.64\pm 1.10\) \(47.00\pm 0.32\) \(47.76\pm 0.26\) \(44.31\pm 0.26\) \(49.90\pm 0.67\) \(50.72\pm 0.55\) \(47.26\pm 0.65\) \(50.75\pm 0.78\) \(51.55\pm 0.62\) \(48.14\pm 0.76\)
MEDAE (s) \(32.70\pm 0.88\) \(35.31\pm 0.81\) \(30.36\pm 0.90\) \(31.13\pm 0.17\) \(31.92\pm 0.30\) \(28.95\pm 0.16\) \(31.92\pm 0.28\) \(32.75\pm 0.44\) \(29.62\pm 0.30\) \(32.06\pm 0.08\) \(33.02\pm 0.27\) \(29.90\pm 0.32\)
  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