Protein Engineering

ube4b · Fully connected

reference_ube4b_fc_contiguousf4_20261104-171023

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
train2.72370.64810.58780.36072870.00000.56991.00001.00006.38919.18450.50000.00000.94011.00004.46879.18450.46000.00000.82650.96003.68769.18450.39000.00000.78010.96493.41479.18450.40940.00000.7954
validation2.4654-0.0693-0.0553-0.08511218.0000-0.00932200.00000.2000-1.77920.69610.00000.0000-0.18600.1400-1.50671.01850.00000.0000-0.16500.1800-1.34191.29310.00000.0000-0.14480.0698-1.63431.01850.00000.0000-0.1903
test3.0355-0.2564-0.2046-0.0901933.0000-0.03682656.00000.0000-2.4776-0.66270.00000.0000-0.26070.0200-3.02850.74610.00000.0000-0.46660.0400-2.95810.87170.00000.0000-0.52290.0263-3.01960.74610.00000.0000-0.4395

Loss per epoch

training validation

Config

name: reference_ube4b_fc_contiguousf4
task: dms
data:
  dataset: ube4b
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: contiguous
  split_fold: 4
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