Table 2 Malisam and Malidup benchmarks

From: Protein remote homology detection and structural alignment using deep learning

 

Malidup

 

Malisam

 

Method

F1 score

Number detected

F1 score

Number detected

BLAST

0.019 ± 0.019

5

0.000 ± 0.000

2

HMMER

0.020 ± 0.02

8

0.020 ± 0.020

3

Needleman–Wunsch

0.098 ± 0.010

234

0.025 ± 0.003

129

Smith–Waterman

0.114 ± 0.010

234

0.031 ± 0.003

129

DeepBLAST

0.265 ± 0.020

234

0.066 ± 0.009

129

Mammoth-local

0.483 ± 0.020

234

0.187 ± 0.017

129

FAST

0.569 ± 0.026

234

0.300 ± 0.030

129

TM-align

0.576 ± 0.024

234

0.393 ± 0.031

129

Dali

0.791 ± 0.014

234

0.619 ± 0.029

129

  1. Sequence and structure alignment methods measured by their F1 score. FAST, TM-align, Dali and Mammoth-local are structure–structure alignment methods and provide a structure-informed upper bound for this benchmark, as many of the most challenging alignments in this benchmark are ultimately structure derived or curated with a structure–structure alignment as an oracle. The best F1 scores for sequence and structure alignment methods are highlighted in bold.
Back to article page