DeepLoc 2.1: multi-label membrane protein type prediction using protein language models

Abstract

DeepLoc 2.0 is a popular web server for the prediction of protein subcellular localization and sorting signals. Here, we introduce DeepLoc 2.1, which additionally classifies the input proteins into the membrane protein types Transmembrane, Peripheral, Lipid-anchored and Soluble. Leveraging pre-trained transformer-based protein language models, the server utilizes a three-stage architecture for sequence-based, multi-label predictions. Comparative evaluations with other established tools on a test set of 4933 eukaryotic protein sequences, constructed following stringent homology partitioning, demonstrate state-of-the-art performance. Notably, DeepLoc 2.1 outperforms existing models, with the larger ProtT5 model exhibiting a marginal advantage over the ESM-1B model. The web server is available at https://services.healthtech.dtu.dk/services/DeepLoc-2.1.

Graphical Abstract

Figure 1.

An example display of the predictions from the web server. All results from the tables can be downloaded as a comma-separated file (CSV) at the top of the page, which includes the predictions for the subcellular localization, membrane protein type and sorting signals. The attention plot and attention values can be downloaded separately. The predicted subcellular localization, membrane protein type and sorting signal labels are listed, along with prediction score tables. The predicted locations and membrane protein types in the tables are highlighted in green, with the intensity of the color indicating the certainty of the prediction. If none of the scores surpasses the threshold, the label closest to its threshold is selected. Elevated values in the logo-like plot indicate important regions in the sequence for the subcellular localization prediction, potentially corresponding to sorting signals. This is intended as a general guideline, and for a more in-depth and precise analysis of these signals, specialized tools like SignalP (14), TargetP (15) or NetGPI (16) can be employed.