Orbitrap Astral mass spectrometry driven data-independent acquisition (DIA) strategy enables deep profiling in shotgun proteomics and is increasingly adopted in single-cell proteomics (SCP). However, the high proportion of missing values reported by existing DIA software remains a bottleneck for sensitive SCP analysis. Here, we present ApuQuant, an Orbitrap Astral DIA data analysis software that performs cohort-level re-identification and quantification. In ApuQuant, we apply a contrastive learning model to Match Between Run (MBR) analysis and introduce a false discovery rate estimation algorithm to rule out false MBR results across runs within a cohort. Compared with DIA-NN and Spectronaut, ApuQuant reduced missing values from 43.45%–50.82% to nearly 1% on a technical-replicate dataset, thereby enabling ultra-sensitive data analysis. Further validation on additional datasets demonstrated that this improvement was achieved without increasing false positive results. Finally, ApuQuant was applied to an A549-single-cell proteomics dataset and showed a 39.99% increase in protein identifications, illustrating a more comprehensive characterization of proteomic heterogeneity at single-cell resolution.