List your company
ResearchFormerly WizWand

Protein fitness regression on Spike-ACE2 low-N setting (192 train, 48 val)

0.488Spearman CorrelationTranception_L
0.324720.367110.40950.45189Oct 30, 2024
Updated 8mo ago

Evaluation results

MethodLinks
Tranception_Ltraining_setting=low-N (192 training / 48 validation samples), head=2-layer MLP, backbone=frozen, epochs=50, batch_size=16, learning_rate=0.001, loss=MSE2024.100.488
MutaPLMtraining_setting=low-N (192 training / 48 validation samples), head=2-layer MLP, backbone=frozen, epochs=50, batch_size=16, learning_rate=0.001, loss=MSE2024.100.481
ConFittraining_setting=low-N (192 training / 48 validation samples), head=2-layer MLP, backbone=frozen, epochs=50, batch_size=16, learning_rate=0.001, loss=MSE2024.100.412
Augmented ESMtraining_setting=low-N (192 training / 48 validation samples), head=2-layer MLP, backbone=frozen, epochs=50, batch_size=16, learning_rate=0.001, loss=MSE2024.100.363
Augmented EVmutationtraining_setting=low-N (192 training / 48 validation samples), head=2-layer MLP, backbone=frozen, epochs=50, batch_size=16, learning_rate=0.001, loss=MSE2024.100.354
Ridge Regressiontraining_setting=low-N (192 training / 48 validation samples), epochs=50, batch_size=16, learning_rate=0.001, loss=MSE2024.100.335
ESM-2training_setting=low-N (192 training / 48 validation samples), head=2-layer MLP, backbone=frozen, epochs=50, batch_size=16, learning_rate=0.001, loss=MSE2024.100.331