本工具用于蛋白跨膜结构预测，统一支持跨膜 α 螺旋与 β-桶体系，不限物种来源。输入支持多条 FASTA 序列，适用于候选蛋白的批量初筛、膜蛋白表达方案评估、以及信号肽与跨膜区联合判断场景。
模型输出包含全局类别判定、残基层级拓扑标签、跨膜区段边界与信号肽切割位点信息。请注意：该工具用于拓扑结构判定，不直接用于亚细胞定位结论。
1. 蛋白质序列（支持10条 FASTA）:
已解析序列数: 0，总残基数: 0
标签说明（残基层级）
- B：Beta Strand (TM)，跨膜 β 折叠链
- I：Inside (Cytoplasm)，膜内侧/胞质侧
- O：Outside (Extracellular)，膜外侧/胞外侧
- M：Alpha Helix (TM)，跨膜 α 螺旋
- P：Periplasm，周质空间区域
- S：Signal Peptide，信号肽区域
模型性能指标
====================================================================================
BENCHMARK REPORT (Filtered Test Set) [Ensemble: 3 models, method=vote]
====================================================================================
Total Residues Evaluated: 2500818
Overall Accuracy: 0.9855
Overall MCC: 0.9188
Balanced Accuracy: 0.9185
Macro Precision/Recall/F1:0.9299 / 0.9185 / 0.9238
Weighted P/R/F1: 0.9875 / 0.9855 / 0.9865
Micro P/R/F1: 0.9875 / 0.9855 / 0.9865
Classes absent in y_true for this benchmark split: P
Mean Sequence Accuracy: 0.9756
Median Sequence Accuracy: 1.0000
Sequence Exact Match: 0.8535
[Per-Protein Classification]
Protein Accuracy: 0.9879
Protein MCC: 0.9621
Protein Macro P/R/F1: 0.8878 / 0.9375 / 0.9025
Type | Support | Precision | Recall | F1
------------------------------------------------------------------
GLOBULAR | 4044 | 0.9948 | 0.9953 | 0.9951
SIGNAL | 273 | 0.9672 | 0.9707 | 0.9689
TM | 566 | 0.9715 | 0.9629 | 0.9672
SP+TM | 15 | 0.5600 | 0.9333 | 0.7000
BETA | 63 | 0.9455 | 0.8254 | 0.8814
[Signal Peptide Metrics]
SP Presence Acc/MCC: 0.9966 / 0.9715
SP Presence P/R/F1: 0.9565 / 0.9904 / 0.9731
CS Exact Accuracy (true SP): 0.8103
CS Relaxed@+/-2 (true SP): 0.9196
CS Exact P/R/F1: 0.7826 / 0.8103 / 0.7962
CS Relaxed P/R/F1: 0.8882 / 0.9196 / 0.9036
[Per-Segment Metrics]
Segment | TP | FP | FN | Precision | Recall | F1 | BoundaryMAE
----------------------------------------------------------------------------------------------------
B | 763 | 68 | 48 | 0.9182 | 0.9408 | 0.9294 | 0.9463
M | 2544 | 1129 | 483 | 0.6926 | 0.8404 | 0.7594 | 1.7750
S | 292 | 38 | 19 | 0.8848 | 0.9389 | 0.9111 | 0.1284
TM | 3307 | 1197 | 531 | 0.7342 | 0.8616 | 0.7929 | 1.5838
[Per-Class Performance (Residue Level)]
Class | Support | Precision | Recall | F1 | MCC | Specificity
------------------------------------------------------------------------------------------------
B | 7210 | 0.8504 | 0.8513 | 0.8508 | 0.8504 | 0.9996
I | 2263185 | 0.9953 | 0.9931 | 0.9942 | 0.9395 | 0.9549
M | 64393 | 0.9172 | 0.8404 | 0.8771 | 0.8749 | 0.9980
O | 158275 | 0.9129 | 0.9433 | 0.9278 | 0.9230 | 0.9939
P | 0 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.9980
S | 7755 | 0.9737 | 0.9645 | 0.9691 | 0.9690 | 0.9999
[Confusion Matrix Counts]
B I M O P S
B 6138 181 0 486 405 0
I 247 2247605 2794 9067 3370 102
M 0 5705 54115 4535 0 38
O 833 4742 2067 149306 1265 62
P 0 0 0 0 0 0
S 0 85 24 165 1 7480
[Confusion Matrix Row-Normalized]
B I M O P S
B 0.851 0.025 0.000 0.067 0.056 0.000
I 0.000 0.993 0.001 0.004 0.001 0.000
M 0.000 0.089 0.840 0.070 0.000 0.001
O 0.005 0.030 0.013 0.943 0.008 0.000
P 0.000 0.000 0.000 0.000 0.000 0.000
S 0.000 0.011 0.003 0.021 0.000 0.965
====================================================================================
最后更新时间：2026-05-25