Peptide

Peptides are chains of amino acids connected by peptide bonds. When these chains are shorter than twenty amino acids, they are known as oligopeptides, which include dipeptides, tripeptides, and tetrapeptides. Longer, continuous chains are called polypeptides, and once a polypeptide reaches a molecular mass of 10,000 daltons or more, it is classified as a protein. Most peptides are linear molecules, featuring a free amine group at one end (the N-terminus) and a carboxyl group at the other (the C-terminus), though macrocyclic peptides form a separate category.

Peptides can be grouped by their source or function, with categories such as plant, bacterial, venom, brain, and cardiovascular peptides, among many others. Some peptides are produced by ribosomes and later modified through proteolysis, often serving as hormones or signaling molecules in higher organisms. Microbes like bacteria also generate peptide antibiotics, such as microcins and bacteriocins. Post-translational modifications are common and include phosphorylation, glycosylation, and disulfide bond formation. While peptides are generally linear, lariat structures have been observed, and more unusual changes—like the conversion of L-amino acids to D-amino acids—occur in platypus venom.

Nonribosomal peptides are built by enzyme complexes rather than ribosomes. A well-known example is glutathione, which helps protect aerobic organisms from oxidative damage. These peptides are especially common in unicellular organisms, plants, and fungi, and are synthesized by modular enzyme systems called nonribosomal peptide synthetases. These complexes often contain multiple modules that perform diverse chemical reactions, frequently producing cyclic or highly complex structures, though linear forms also exist. Because this machinery is related to that used for fatty acid and polyketide synthesis, hybrid compounds are common, and the presence of oxazoles or thiazoles often indicates a nonribosomal origin.

Peptones are produced by digesting animal milk or meat with proteolytic enzymes. Besides small peptides, they contain fats, metals, salts, vitamins, and other biological compounds, and are used in nutrient media for growing bacteria and fungi. Peptide fragments are pieces of proteins used to identify or quantify the source protein; they often come from lab-controlled enzymatic degradation, but can also arise from natural degradation in forensic or paleontological samples.

Peptides interact with proteins and other macromolecules, playing key roles in human cell signaling and immune modulation. Studies suggest that 15–40% of all protein–protein interactions in human cells involve peptides, and peptide-based products account for at least 10% of the pharmaceutical market.

Machine learning and deep learning methods are widely used to classify, screen, and design peptides based on sequence or structural data. These computational approaches are especially useful when experimental screening is too costly, slow, or difficult to scale. A typical workflow includes dataset curation, converting peptide sequences or structures into numerical features, model optimization, and validation. Common representations include amino acid composition, physicochemical descriptors, substitution matrices, and embeddings from protein or peptide language models. These techniques have been applied to antimicrobial, cell-penetrating, and anticancer peptides, though challenges remain with dataset biases, inconsistent benchmarks, and interpretability of complex models.

The chemical space of peptides is a multidimensional landscape defined by molecular descriptors or fingerprints, where the distance between molecules indicates chemical or functional similarity. This space can be mapped using primary sequences, three-dimensional structures, or both. Key properties for mapping include molecular weight, lipophilicity (logP and logD), topological polar surface area (TPSA), and hydrogen-bond dynamics. Dimensionality-reduction techniques like PCA, t-SNE, and UMAP, along with clustering algorithms, help visualize peptide libraries and identify groups with related activities. Peptides differ from traditional small molecules due to their residue sequence, flexible amide backbone, and susceptibility to chemical modifications, all of which influence bioavailability and membrane permeability. Computational analysis uses notation systems such as FASTA, HELM, and BILN to encode both canonical and modified sequences, with modifications like cyclization being common.