Protein Engineering

gb1 · Fully connected

reference_gb1_fc_modulof2_20261104-163208

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
train0.51150.95750.96150.906734843.00000.74861404.00001.00001.60581.82200.20000.00000.95471.00001.48391.82200.38000.00000.96401.00001.44921.82200.46000.00000.97160.99841.07902.01760.62160.00000.9727
validation4.5823-0.0953-0.1066-0.04484705.00000.10796916.00000.1000-1.60430.23640.00000.00000.07660.1200-1.57690.53170.00000.00000.08940.1100-1.78520.53170.00000.00000.03460.1400-1.76521.32510.00000.00000.0414
test5.2660-0.1869-0.1536-0.18591545.00000.10985314.00000.0000-5.4456-4.33480.00000.0000-0.71660.0000-5.2678-0.30690.00000.0000-0.69760.0000-5.2450-0.16300.00000.0000-0.70480.0100-5.15920.27200.00000.0000-0.6960

Loss per epoch

training validation

Config

name: reference_gb1_fc_modulof2
task: dms
data:
  dataset: gb1
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: modulo
  split_fold: 2
model:
  name: fcn
  hidden:
  - 1000
  dropout: 0.2
train:
  epochs: 300
  batch_size: 64
  learning_rate: 0.0001
  patience: 15
  min_delta: 1.0e-05
  seed: 7