Česká společnost pro hmotnostní spektrometrii

(13. ročník České konference hmotnostní spektrometrie a 11. Neformální proteomické setkání - ThS-10)
Human IMPs: A tool for comparing outputs of different tools for transmembrane segment predictions in the human proteome

Ondřej Vít ^1,2 *, Tereza Hrdinová ¹, Jiří Petrák ¹

1. BIOCEV, First Faculty of Medicine, Charles University, Vestec
2. Institute of Medical Biochemistry and Laboratory Diagnostics, General University Hospital, Prague

Abstrakt

More than a quarter of human genes encode integral membrane proteins (IMPs), which perform key biological functions and represent the majority of targets for currently approved drugs. IMPs are defined by the presence of transmembrane regions, typically alpha helices. Accurately identifying these regions remains challenging, as available algorithms often provide different results and require computational skills for batch processing. Meanwhile, UniProt annotation data, though easily accessible, are obtained from diverse sources and tend to differ from computational predictions.
To facilitate comparison of prediction outputs and UniProt annotations, we developed Human IMPs, a web application integrating transmembrane helix data from UniProt and predictions from four widely used algorithms: Phobius, DeepTMHMM, TMBED, and TOPCONS. The application features interactive graphical visualization and supports batch analysis of multiple proteins simultaneously.
Human IMPs uses the UniProt API to fetch curated annotation data across the human proteome. Predictions are stored in an optimized database, enabling rapid retrieval and comparison. The graphical interface displays predicted transmembrane regions as colored bars aligned to protein sequences, making discrepancies immediately apparent. Batch processing allows analysis of large protein datasets with results exportable in CSV or TSV formats. We demonstrate these capabilities by comparing prediction models on IMPs identified in pheochromocytoma and paraganglioma tumor proteomes.
Future development plans include integration of additional prediction algorithms and implementation of consensus transmembrane region computation combining multiple methods for more reliable predictions. Explore Human IMPs at hwllffrdd.cz/imps/.

* Korespondující autor: ondrvit@gmail.com

Poděkování:

Supported by MZČR (NU23-01-00323), Charles University (SVV 260763) and EXCELES LX22NPO5102.