Cancer neoantigen prioritization through sensitive and reliable proteogenomics analysis

Area of Science:

• Computational Biology
• Proteomics
• Genomics

Background:

• Genomics-based neoantigen discovery can be improved with proteomic data.
• Current quality control methods for variant peptide identification in proteogenomics lack consensus.
• Accurate retention time prediction is crucial for peptide identification.

Purpose of the Study:

• To develop and validate a novel metric for evaluating quality control methods in proteogenomics.
• To introduce AutoRT, a deep learning algorithm for accurate peptide retention time prediction.
• To provide practical guidance for selecting quality control strategies to enhance neoantigen discovery.

Main Methods:

• Developed AutoRT, a deep learning algorithm for predicting peptide retention times.
• Utilized the difference between predicted and observed retention times as a quality control metric.
• Analyzed three cancer datasets (287 tumor samples) using various quality control strategies.
• Implemented the recommended strategy in the NeoFlow computational workflow.

Main Results:

• AutoRT demonstrated high accuracy in retention time prediction.
• Different quality control strategies yielded significantly different numbers of identified variant peptides and neoantigens.
• The retention time metric provided a systematic way to evaluate and compare quality control methods.
• The NeoFlow workflow demonstrated enhanced sensitivity for putative neoantigen discovery.

Conclusions:

• Retention time prediction accuracy is a valuable metric for assessing peptide identification quality control in proteogenomics.
• The proposed method and AutoRT algorithm offer practical guidance for optimizing neoantigen discovery pipelines.
• NeoFlow facilitates more sensitive identification and prioritization of putative neoantigens for cancer immunotherapy research.