Protein Engineering

ube4b · Fully connected

reference_ube4b_fc_plm_contiguousf0_20261104-175827

Metrics

Setmsepearsonrspearmanrr2above_wt_nabove_wt_spearmanrbudget_to_maxhit_rate_at_10mean_at_10max_at_10precision_at_10tie_share_at_10captured_at_10hit_rate_at_50mean_at_50max_at_50precision_at_50tie_share_at_50captured_at_50hit_rate_at_100mean_at_100max_at_100precision_at_100tie_share_at_100captured_at_100hit_rate_at_1pctmean_at_1pctmax_at_1pctprecision_at_1pcttie_share_at_1pctcaptured_at_1pct
train2.93600.68110.63200.29684241.00000.553215.00001.00005.57237.74370.20000.00000.82431.00005.05209.18450.38000.00000.87111.00004.56289.18450.43000.00000.86901.00004.15129.18450.52110.00000.8905
validation3.25350.29660.32480.0832458.0000-0.08582006.00000.5000-0.20561.03880.00000.00000.25760.5200-0.08523.46860.04000.00000.34800.4200-0.20314.42140.10000.00000.36600.5000-0.06443.46860.03850.00000.3178
test2.55430.19230.1647-0.12431218.00000.0126135.00000.4000-0.49371.37420.00000.00000.06220.4400-0.27981.87170.00000.00000.12810.3700-0.35792.25020.01000.00000.12610.4419-0.32041.87170.00000.00000.1153

Loss per epoch

training validation

Config

name: reference_ube4b_fc_plm_contiguousf0
task: dms
data:
  dataset: ube4b
  encoding: one_hot,aa_index,plm,plm_sum
  split_dir: null
  split_scheme: contiguous
  split_fold: 0
model:
  name: fcn
  hidden:
  - 100
  - 100
  - 100
  dropout: 0.2
train:
  epochs: 300
  batch_size: 64
  learning_rate: 0.0001
  patience: 15
  min_delta: 1.0e-05
  seed: 7