DAS-TMfilter

DAS-TMfilter predicts transmembrane (TM) segments in protein sequences using a Dense Alignment Surface (DAS)-based comparison to improve selectivity and reduce false positives.

Key Features:

• Dense Alignment Surface (DAS) scoring: Uses low-stringency dot-plots and a specialized scoring matrix to compare query sequences against a library of non-homologous membrane proteins.
• Hydrophobicity profiling: Generates a high-precision hydrophobicity profile from DAS comparisons to identify potential TM segments.
• Second prediction cycle: Implements an additional prediction pass specifically aimed at reducing false positive TM region calls.
• Comparative evaluation against documented TM segments: Compares the query sequence to a library of documented transmembrane protein segments as part of the second cycle.
• Empirical threshold classification: Applies an empirically determined threshold on the comparative performance to classify sequences as non-transmembrane when appropriate.
• High sensitivity: Achieves ~95% success rate in identifying TM segments within a learning set of 128 documented transmembrane proteins.
• High selectivity: Achieves ~99% accuracy over a non-redundant set of 526 soluble proteins with known three-dimensional structures by eliminating many falsely predicted single-pass membrane proteins.

Scientific Applications:

• Transmembrane segment detection and annotation: Identification and annotation of helical TM regions in protein sequences.
• Discrimination of membrane versus soluble proteins: Reduces false positives when distinguishing integral membrane proteins from soluble proteins.
• Protein structure–function and membrane protein research: Supports studies of protein structure–function relationships and membrane protein analysis.

Methodology:

Applies the Dense Alignment Surface (DAS) method with low-stringency dot-plots and a specialized scoring matrix against a library of non-homologous membrane proteins to generate hydrophobicity profiles, then performs a second comparative cycle against a library of documented transmembrane segments and uses an empirically determined threshold to classify non-transmembrane sequences.