TMSEG

TMSEG predicts transmembrane proteins and their transmembrane helices (TMHs) and infers their membrane topology to support identification and characterization of TMPs for structural biology and drug discovery.

Key Features:

• Machine Learning Integration: Employs machine learning techniques combined with empirical filters to classify proteins and improve prediction accuracy.
• Helical TMP Detection Sensitivity and False Positive Rate: Achieves a sensitivity of 98 ± 2% for identifying helical TMPs with a false positive rate of 3 ± 1% on the reported test set.
• TMH Prediction Precision and Recall: Predicts individual TMHs with precision of 87 ± 3% and recall of 84 ± 3%, and correctly predicts placement and inside/outside topology of all TMHs in 63 ± 6% of helical TMPs.
• Reduction in Misclassifications: Reports 200 to 1600 fewer misclassifications than the second and third best methods in human datasets, and 4400 fewer mistakes than a simple hydrophobicity-based method.
• Add-on Improvement Capability: Can be used to enhance the accuracy of existing transmembrane prediction methods.

Scientific Applications:

• Structural Biology: Supports prediction of TMPs and TMHs where experimental structure determination of membrane proteins is challenging.
• Drug Discovery: Facilitates identification and topology characterization of membrane protein drug targets to aid targeted therapy development.

Methodology:

Applies machine learning techniques combined with empirical filters for classification and TMH/topology prediction and was evaluated on a non-redundant dataset of 41 TMPs and 285 soluble proteins.