T2Pdecoder enables protein-centric analyses from transcriptomic data.

Area of Science:

• Computational biology
• Genomics
• Proteomics

Background:

• Protein quantification is less extensive than RNA quantification, particularly in IDH-mutant gliomas.
• Predicting protein levels from RNA is challenging due to weak correlation but valuable for biological insights.
• Existing methods predict limited protein subsets, restricting proteomic applications.

Purpose of the Study:

• To develop an integrative multi-omics deep learning model, T2Pdecoder, for predicting broad protein abundance profiles from RNA data.
• To leverage shared embedding spaces between protein and RNA for improved prediction accuracy.
• To enable protein-centric analyses from transcriptomic data in cancer research.

Main Methods:

• T2Pdecoder utilizes a deep learning approach integrating multi-omics data.
• The model learns a shared embedding space between RNA and protein expression data.
• Evaluation performed on diverse glioma datasets, comparing against RNA-only baselines.

Main Results:

• T2Pdecoder showed modest but consistent improvements in predicting protein abundance compared to RNA-only methods.
• The model more accurately recapitulated protein-level pathway enrichment patterns.
• Applications revealed functional glioma subgroups with survival differences and identified cell markers from single-cell RNA data.

Conclusions:

• T2Pdecoder facilitates protein-centric analyses using transcriptomic data.
• The model offers complementary biological insights beyond conventional RNA-only analyses in cancer research.
• T2Pdecoder aids in reducing batch effects in single-cell RNA data and identifying cell markers.