DeepSCP: utilizing deep learning to boost single-cell proteome coverage.

Area of Science:

• Biochemistry
• Computational Biology
• Proteomics

Background:

• Multiplexed single-cell proteomes (SCPs) quantification via mass spectrometry enhances coverage but faces limitations in protein identification.
• Computational methods offer potential to significantly increase protein identifications in SCP analysis.

Purpose of the Study:

• To introduce DeepSCP, a novel deep learning framework designed to boost SCP coverage and protein identification.
• To improve the accuracy of peptide-spectrum match (PSM) predictions using advanced computational techniques.

Main Methods:

• DeepSCP utilizes deep learning to predict retention times across multiple SCP sample sets and fragment ion intensities.
• An optimized-ensemble learning model is employed for accurate PSM label prediction.
• The framework was evaluated on public and in-house SCP datasets using the target-decoy competition method.

Main Results:

• DeepSCP demonstrated superior performance compared to existing state-of-the-art methods in enhancing SCP identification.
• The framework successfully identified more confident peptides and proteins, with a controlled q-value of 0.01.
• DeepSCP provides a convenient and low-cost computational solution for boosting protein identification.

Conclusions:

• DeepSCP significantly enhances single-cell proteome identification by improving PSM accuracy through deep learning.
• The framework facilitates the advancement and application of single-cell proteomics.
• DeepSCP offers a valuable tool for researchers aiming to increase protein coverage in complex proteomic datasets.