Recovering Sparse Low-rank Blocks in Mass Spectrometry∗

We develop a novel sparse low-rank block (SLoB) signal recovery framework that simultaneously exploits sparsity and low-rankness to accurately identify peptides (fragments of proteins) from biological samples via tandem mass spectrometry (TMS). To efficiently perform SLoB-based peptide identification, we propose two novel recovery algorithms, an exact iterative method and an approximate greedy algorithm, and provide analytical recovery guarantees. Using experiments with synthetic and real-world TMS data, we demonstrate that the proposed framework and algorithms are capable of substantially outperforming existing sparse signal recovery techniques.