Extended Data Fig. 2: Sequence features of the archaeasins.
From: Deep learning reveals antibiotics in the archaeal proteome
(a) Bidimensional sequence space visualization of peptide sequences found in DBAASP, all predicted antimicrobial EPs discovered by APEX, and peptides that were synthesized and validated in this study. Uniform Manifold Approximation and Projection (UMAP) was used to reduce the feature representation to two dimensions for visualization purposes. (b) Frequency of the gene ontology annotations of the top 265 archaeasins with mean MIC < 80 μmol L−1. (c) Amino acid frequency in archaeasins compared with known AMPs from DBAASP, APD3, and DRAMP 3.0 databases, and other encrypted peptides from the human proteome discovered by APEX6 and a scoring function4. (d) To further explore differences at the individual residue level, we calculated the percentage of each amino acid residue within every peptide sequence and used a one-sided Mann-Whitney U test to compare these percentages between peptide sets. A significance threshold of p < 0.01 was used to identify amino acids that were significantly more or less abundant between specific sets. The exact p-values for each comparison are shown in Supplementary Data 3 and 4 files (under Data availability). (e-g) Relative abundance of the amino acid content of Archaea encrypted peptides identified by APEX (red) and (d) AMPs from the DBAASP, APD3, and DRAMP 3.0 database (blue), (e) human proteome encrypted peptides discovered by APEX (orange), and (f) human encrypted peptides found by the scoring function (yellow). The frequency of amino acid was normalized by the total number of amino acid residue counts. Chi-square test of independence of variables in a contingency table was used to compare the amino acid composition in c-f, p values were 0, that is, below machine precision levels suggesting that they are statistically significant.
