Protein Engineering

ube4b · Fully connected

reference_ube4b_fc_contiguousf3_20261104-162352

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
train2.61750.58830.54230.32595445.00000.47117213.00001.00005.53358.62460.30000.80170.98004.40948.62460.20000.77630.95003.89118.62460.34000.75680.95003.80488.62460.47270.8195
validation2.8439-0.1265-0.1207-0.0213933.00000.00512063.00000.4000-1.29422.06240.00000.0000-0.03730.1800-1.74982.06240.00000.0000-0.15780.2600-1.56173.49970.02000.0000-0.13060.1579-1.81232.06240.00000.0000-0.1636
test4.3991-0.1320-0.1176-0.1720323.0000-0.09472388.00000.0000-2.8535-0.93570.00000.0000-0.12730.0000-2.8503-0.14850.00000.0000-0.15960.0200-2.75790.77860.00000.0000-0.16540.0000-2.9990-0.93570.00000.0000-0.1672

Loss per epoch

training validation

Config

name: reference_ube4b_fc_contiguousf3
task: dms
data:
  dataset: ube4b
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: contiguous
  split_fold: 3
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