Fig. 5: Searching complex proteomes acquired on the Orbitrap Astral with fully predicted spectral libraries.

a, Six replicates of 200-ng HeLa bulk data were analyzed on the Orbitrap Astral with a 60-SPD (21 min) gradient. A fully predicted alphaPeptDeep library was used for a two-step search in alphaDIA. Different search engines were used for comparison. Evosep liquid chromatography illustration created with BioRender. b, Mean precursors identified across search engines (n = 6) c, Mean modified peptides identified across processing methods (n = 6) d, Protein groups identified at given CV cutoffs. e, Analysis time for different processing steps when analyzed with on a 32-core machine. f, Arabidopsis entrapment search using the fully predicted library workflow. The share of identified Arabidopsis proteins at 1% target–decoy FDR is shown. g, Venn diagram showing the overlap of proteotypic peptides across processing methods. h, Analysis of protein overlap between processing methods. Peptides were mapped back to the same reference proteome, discarding ambiguous matches. The median number of peptides per protein is shown. i, Mixed-species experiment for establishing quantitative accuracy. Human, yeast and E. coli proteomes were combined in defined ratios and protein ratios are shown for proteins quantified in at least three of five replicates (box plot defined as per Methods).