Protein Engineering

ube4b · Fully connected

reference_ube4b_fc_contiguousf1_20261104-154553

Metrics

Setmsepearsonrspearmanrr2above_wt_nabove_wt_spearmanrbudget_to_maxhit_rate_at_10mean_at_10max_at_10precision_at_10captured_at_10hit_rate_at_50mean_at_50max_at_50precision_at_50captured_at_50hit_rate_at_100mean_at_100max_at_100precision_at_100captured_at_100hit_rate_at_1pctmean_at_1pctmax_at_1pctprecision_at_1pctcaptured_at_1pct
train2.33210.53600.53480.26139072.00000.2658157.00000.90003.33736.15800.20000.59310.94003.08486.15800.14000.64670.97003.06577.55330.15000.70130.94842.58027.62880.39200.7623
validation5.88920.03270.0149-0.015488.00000.1050287.00000.1000-2.74971.66520.00000.0000-0.18360.2000-1.96314.16940.10000.0000-0.07960.1900-1.92574.16940.24000.0000-0.09910.0000-3.0158-1.04090.00000.0000-0.2015
test3.7484-0.0923-0.1079-0.0562458.00000.10831239.00000.1000-1.85542.87170.00000.0000-0.03570.0600-2.14902.87170.02000.0000-0.10970.0700-2.03152.87170.01000.0000-0.09510.0769-2.36582.87170.03850.0000-0.1422

Loss per epoch

training validation

Config

name: reference_ube4b_fc_contiguousf1
task: dms
data:
  dataset: ube4b
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: contiguous
  split_fold: 1
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