The primary structures of peptides, originating from food proteins, affect their taste. Connecting primary structure to taste, however, is difficult because the size of the peptide sequence space increases exponentially with increasing peptide length, while experimentally-labeled data on peptides' tastes remain scarce. We propose a method that coarse-grains the sequence space to reduce its size and systematically identifies the most common coarse-grained residue patterns found in known bitter and umami peptides. We select the optimal patterns by performing extensive out-of-sample tests. The optimal patterns better represent the bitter and umami peptides when compared against baseline peptides, bitter peptides with all hydrophobic residues and umami peptides with all negatively charged residues, and peptides with randomly-chosen residues. Our method complements quantitative structure-activity relationship methods by offering generic, coarse-grained bitter and umami residue patterns that can aid in locating short bitter or umami segments in a protein and in designing new umami peptides.
I. INTRODUCTIONSpecial compounds trigger specific tastes: sodium chloride (salty), sugars (sweet), acids, phenols, and alkaloids (bitter), glutamic acids and nucleotides (umami). Tastes are crucial because the gustatory system, the sensory system that helps in perceiving taste, often informs us about safe and harmful foods through their tastes 1 . Further, taste determines most of our food preferences 2 . For example, vegetables such as cabbage, cucumber, or spinach, often taste bitter since they contain plant alkaloidswhich can be toxic if consumed in large amounts and are known to have excessive bitter taste 2,3 -and, consequently, we avoid them.Bitter and umami represent two major taste modalities that peptides commonly have; the third one is sweet 4 . Interestingly, while salty, sour, sweet, and bitter were recognized as basic tastes early on, umami was recognized as the fifth basic taste only around the beginning of this century when umami taste receptors were identified [5][6][7] . As a result, the study of umami peptides is a more recent endeavor compared to, for example, the study of bitter peptides 8,9 .Bitter peptides are often found in fermented foods 10,11 and protein hydrolysates 3 , while umami peptides are found in savory foods such as parmesan cheese, fermented soy sauce, and seaweed 8 . As we tend to avoid bitter foods and seek savory ones, classifying foods based on the taste responses they evoke and modulate, and finding the physicochemical reasons causing those responses are indispensable steps in designing new nutritional and palatable foods. The growing number of curated databases, and analysis tools, of bitter-and umamia)