Hydrolyzing protein is the most common method for preparing antioxidant peptides. Acid-base hydrolysis uses acid or alkali to hydrolyze protein. Some researchers hydrolyzed ostrich egg white protein with 0.25 mol/L NaOH at 40 ºC for eight hours. They proved that the antioxidant activity of the protein was significantly enhanced after hydrolysis (Khueychai et al., 2019). This method is cheap and time-saving. Nonetheless it is difficult to control the degree of hydrolysis and product quality, because the violent reaction conditions involved are destructive to amino acids (Hall and Ahmad, 1997).

Microbial fermentation can produce enzymes and enzymatic hydrolysis simultaneously. This method also can directly isolate and purify antioxidant peptides (Chai et al., 2020). After fenugreek protein was fermented with Lactococcus lactis for 24 hours, the free radical scavenging activities of 1,1-diphenyl-2-picrylhydrazyl (DPPH) and 2,2’-azinobis (3-ethylbenzothiazoline-6-sulfonic acid) ammonium salt (ABTS) increased by 42.9 and 40%, respectively (Setti et al., 2017). However, the process of screening microorganisms is cumbersome and the safety of microbial enzyme production is uncontrollable (He et al., 2012).

Enzymatic hydrolysis uses proteolytic enzymes to hydrolyze protein into small molecular polypeptides. It is characterized by mild conditions, strong operability, strong specificity and fewer by-products. The same protein hydrolyzed by different enzymes produces different antioxidant peptides (da Rosa et al., 2018). A study has measured the antioxidant activity of camel colostrum protein hydrolyzed by pepsin, trypsin, pancreatin and another two enzymes. The results showed that the proteolytic peptide hydrolyzed by pancreatin had the highest antioxidant activity (Oussaief et al., 2019). Hydrolysis conditions are also critical. Some researchers have optimized the hydrolysis conditions from egg white with chymotrypsin and pepsin through response surface methodology, and the antioxidant activity of the peptide was improved (Yuan et al., 2020). Table 1 summarizes experiments related to the preparation of antioxidant peptides by enzymatic hydrolysis.

However, enzymatic hydrolysis usually obtains crude peptides, so it needs to be separated and purified. Usually, ultrafiltration or gel chromatography can be carried out according to the size of antioxidant peptides. Ion exchange chromatography and reverse chromatography can also be used (based on the charge and hydrophobicity of the antioxidant peptide, respectively) (Vanvi and Tsopmo, 2016). After that, tandem mass spectrometry is applied to identify the antioxidant peptides. One study extracted antioxidant peptide from duck plasma hydrolysate (DPH) and then separated and purified the antioxidant peptide by ultrafiltration, dimensional exclusion chromatography and reversed-phase high performance liquid chromatography. The researchers also identified the DPH fraction with the highest antioxidant activity by nanoscale liquid chromatography-tandem mass spectrometry (Yang et al., 2020). In fact, DPPH, ABTS free-radical scavenging and oxygen-radical absorbance capacity (ORAC) tests are generally used to evaluate the effect of antioxidant peptides (Kim et al., 2018). Other researchers have pre-treated the protein with ultrasound, microwave and irradiation to improve the antioxidant activity of the peptide (Wang et al., 2019; Zhang et al., 2019). However, the single use of enzymatic hydrolysis is time consuming when selecting hydrolytic enzymes. The process of identifying antioxidant peptides is complicated and the target antioxidant peptides may not be obtained (Wen et al., 2020). Therefore, bioinformatics has attracted people’s attention.

Bioinformatics based on computer technology can guide the selection of precursor proteins and simulate enzymatic hydrolysis (Zhou et al., 2019). Moreover, the function of the enzymatic hydrolysis products can be predicted and identified based on the sequence, structure and other parameters of proteins and peptides by bioinformatics. It is widely used in the research of hydrolyzing protein to obtain antioxidant peptides (Tejano et al., 2019). This method is often used with databases such as Peptide Ranker, BIOPEP-UWM and UniProtKB (Chen et al., 2017; Pearman et al., 2020). Some commonly used bioinformatic databases are listed in Table 2. These databases contain the amino acid sequences of various proteins and peptides to improve the efficiency of obtaining target peptides at a low cost (Tadesse and Emire, 2020). Darewicz et al. (2016) retrieved the amino acid sequence, molecular weight and chain length of Cyprinus carpio protein from UniProtKB. After using the ClustalW2‒Multiple Sequence Alignment program to eliminate the same sequence, they selected 33 carp protein amino acid sequences with less than 90% identity for further analysis. The number of amino acid residues in the analyzed carp protein amino acid sequence ranged from 62 (light myosin, Q90335) to 1938 (myosin heavy chain, Q2HX56).

These bioinformatic methods can also simulate protease hydrolysis. In 2018, researchers used Peptide Cutter to simulate 100 tripeptides from the myosin of Mizuhopecten yessoensis by pepsin and trypsin (Yu et al., 2018). For example, the BIOPEP-UWM database can not only simulate hydrolysis, but also predict the bioactive peptides released from protein sequences and their bioactivity and physicochemical properties (Yang et al., 2017). Researchers estimated the solubility of antioxidant peptides from flaxseed proteins in water by using Innovagen Peptide Solubility Calculator Proteomic. They also predicted the release of antioxidant peptides by opening the ‘search for active fragments tab of BIOPEP. Thereafter, they estimated several properties and indices of the silico-derived antioxidant peptides by using the application of the ‘Peptides’ package in R or ProtParam of ExPASy. These properties included amino acid composition, amino acid length, molecular weight (MW), isoelectric point (pI), Boman index, net charge and hydrophobicity index (Ji et al., 2019). In another research, Salim and Gan (2020) used Peptide Cutter to simulate the digestion of egg white ovalbumin, then Peptide Ranker was used to screen the obtained peptide sequences. After that, they predicted the peptide binding sites by Pepsite2 and finally used an in vitro assay to verify the bioactivity of the peptide. The AnOxPePred, a method for predicting the free-radical scavenging ability and chelating properties of antioxidant peptides was developed. When constructing the predictive variable, a standard database consisting of peptides and their free-radical scavenging capabilities and chelating properties was created (Olsen et al., 2020).

Although bioinformatics is an effective method to study and predict the antioxidant peptides released by proteins and their potential activities, there are some limitations to this approach. For example, there is a lack of research on the protein sequence in simulated hydrolysis and its influence on enzymatic hydrolysis (Zhou et al., 2019). Bioinformatics simulates the hydrolysis of the protein sequence completely, but it is often necessary to consider the secondary and tertiary structures of protein and other factors such as inhibitors, pH and temperature when verifying in vitro (Bechaux et al., 2020). Some scientists showed that in the peptides obtained by computer simulation of a porcine actin sequence, only a few could be verified through laboratory experiments (Keska and Stadnik, 2016). This proved that bioinformatics was an auxiliary method for selecting protein families and optimal hydrolyses. Bioinformatics was still theoretical, so it needed to be verified by in vitro experiments.

In order to efficiently and economically prepare antioxidant peptides, more and more researchers combine enzymatic hydrolysis and bioinformatics. The amino acid sequence of the target protein can be searched in the database to facilitate the selection of hydrolases, and it is estimated that there are more than 8 million amino acid sequences of proteins in the database (Agyei et al., 2018). Some researchers used the amino acid sequence of the 11S coffee globulin (the entry Q9ZNY2_COFAR and accession number NCBI 13443) deposited in the UniProtKB database. Subsequently, they selected enzymes (pepsin (EC 3.4.23.1), trypsin (EC 3.4.23.4) and chymotrypsin (EC 3.4.21.1)) to simulate gastrointestinal digestion. They carried out the prediction of peptides generated through simulation by using the BIOPEP-UWM “enzyme action”. Based on the results for the prediction of bioactivity of the 11S globulin, a series of in vitro studies was carried out, including protein extraction, enzymatic hydrolysis, protein isolation and determination of its antioxidant capacity. Finally, they proved that the hydrolyzed peptides of roasted coffee beans have higher antioxidant activity (Ribeiro et al., 2021). Computer simulation of hydrolysis can also greatly improve the screening efficiency of hydrolytic enzymes. Jie et al. (2020) used gluten as the bioinformatics template of Caragana korshinskii seed protein, then carried out virtual enzymatic hydrolysis with the assistance of the BIOPEP-UWM database. They determined that papain was the main protease to hydrolyze the Caragana korshinskii seed protein, which improved preparation efficiency.

After simulated hydrolysis, the computer generates an amino acid sequence. By comparing the sequence with the antioxidant peptide sequences reported in databases such as Swiss-Prot, Peptide Ranker or in the literature, the activity of the antioxidant peptide is predicted, evaluated and the target peptide is selected (Tu et al., 2018). For example, Sun et al. (2021) first prepared defatted collagen hydrolysates from yak bone, and then separated and purified the hydrolysate by ultrafiltration and reversed-phase high-performance liquid chromatography. After that, they determined the antioxidant activities of different hydrolysates and screened the best components for identification by mass spectrometry. Mass spectrometry was used to identify components with better antioxidant activity- Swiss-Prot was used to determine the amino acid sequence of the target peptide. Finally, they selected peptides with molecular weight of less than 1 kDa for synthesis according to the possible structural features of antioxidant peptides and the number of enzyme-cutting sites contained in the peptides. Other researchers isolated and purified antioxidant peptides from Asparagus by-products by reverse-phase chromatography, and then identified the peptides by tandem mass spectrometry. Peptide identification was carried out by using the Swiss-Prot database. PeptideRanker was used to predict the probability of the antioxidant activities of these peptides. This study only considered peptides with probability scores higher than 0.95. Therefore, only five peptides were retained and five novel peptide sequences were synthesized, and their antioxidant activity was finally evaluated in vitro (Montone et al., 2019). It is also necessary to predict the toxicity of antioxidant peptides. Harvian et al. (2019) used ToxinPred to analyze the antioxidant peptide obtained from jack bean (Canavalia ensiformis) canavalin protein. ToxinPred is a in silico database for the prediction of the toxicity of selected peptides. The support vector machine (SVM)-based prediction method and SVM threshold value (0.0) were used to separate toxic from non-toxic peptides. Their results showed that the antioxidant peptide was non-toxic. Of course, in vivo or in vitro tests are required to further determine the physiological activity of antioxidant peptides eventually.

In summary, the bioinformatics method can simulate protein hydrolysis, determine the most suitable hydrolase and screen out peptides with good antioxidant effects by comparing the amino acid sequences of peptides gained through experiments with those in the database for in vitro verification. The combination of the two methods will simplify the cumbersome experimental process of traditional enzymatic hydrolysis, save cost, improve preparation efficiency and provide a promising method to improve not only the availability but also the high-value utilization of antioxidant peptides (Ji et al., 2019).

Free radicals exhibit strong reactivity, so they are initiators in oxidation reactions (Tomycz et al., 2011). At present, it is generally recognized that an important cause of oil automatic oxidation is the chain reaction initiated by free radicals, such as hydroxyl radicals, peroxy free radicals and alkoxy free radicals (Sonklin et al., 2018). Hydroxyl radicals can remove the hydrogen atoms in the adjacent fat chain and induce oxidation chain reactions. High-energy peroxy free radicals and alkoxy free radicals can combine hydrogen atoms from adjacent fatty chains to destroy fatty acids and deteriorate oil. Therefore, it is necessary to combine free radical scavengers with free radicals to slow down the rate of chain reactions. Many antioxidant peptides are free radical scavengers. As shown in Scheme 2, together with amino acids, they act as hydrogen donors or electron donors to terminate the chain reactions of free radicals by converting them into stable products (Yarnpakdee et al., 2015).

Scheme 2. Pathways of antioxidant peptides scavenging free radicals. B:H stands for antioxidant peptide. Hydrogen donation: the antioxidant peptide transfers H· to R· to form a stable compound R:H, while the antioxidant peptide converts to B·, which is less likely to initiate a new free radical chain reaction. Electron donation: electron transfer and proton transfer are required (Liang and Kitts, 2014).

The antioxidant capacity of peptides and their amino acid composition are closely related, and the mechanism of amino acids depends on the functional groups of their side chains (Elias et al., 2008). For example, peptide sequences with hydrophobic amino acids, aromatic amino acids and sulfur-containing amino acids usually exhibit strong antioxidant activity (Wu et al., 2015a). Antioxidant peptides containing hydrophobic amino acids such as alanine, valine and leucine have strong antioxidant capacity. The non-polar aliphatic hydrocarbon side chains of these amino acids can enhance the interaction between fatty acids and peptides to prevent hydrogen atoms from being attacked (as shown in Figure 1A) (Rajapakse et al., 2005). Their presence can also increase the solubility of antioxidant peptides in oil, thus promoting interactions with free radicals and exhibiting higher antioxidant activity (Aguilar-Toalá and Liceaga, 2020). Some scientists proposed that hydrophobic amino acids such as valine and aspartic acid present in the peptide sequence could contribute to improving the antioxidant activity of peptides (Najafian and Babji, 2014).

Antioxidant peptides with aromatic amino acids such as tyrosine, tryptophan and phenylalanine in the sequence also have strong antioxidant properties. Aromatic amino acids can not only enhance the solubility of antioxidant peptides in oil like hydrophobic amino acids, but can also scavenge free radicals produced by oxidation by removing a H· from the phenolic hydroxyl or indole group in these amino acids. These amino acids are converted into relatively stable free radicals to terminate the chain reaction and inhibit the automatic oxidation of oil (as shown in Figure 1B) (Wang et al., 2014). One study isolated two peptides containing tryptophan and tyrosine from the hydrolysate of Sardinella aurita, both of which have high DPPH scavenging activity (Bougatef et al., 2010). A related study also showed that the antioxidant activity of the dipeptides Tyr-Leu and Phe-Tyr derived from Perilla is related to the presence of tyrosine (Yang et al., 2018).

Methionine and cysteine are two kinds of sulfur-containing amino acids with active sulfhydryl in their structures. They can be used as precursors for the synthesis of glutathione. The methionine and cysteine in the antioxidant peptide sequence contribute to scavenging free radicals through the electron transfer pathway (Atmaca, 2004). The sulfhydryl group loses electrons to form free radical cations. The free radical cations then undergo a proton transfer reaction to generate stable free radicals and hydrogen ions (Scheme 3).

Scheme 3. Mechanism of scavenging free radicals by sulfur-containing amino acids.

Jiang et al. (2014) obtained His-Asp-His-Pro-Val-Cys and His-Glu-Lys-Val-Cys from Decapterus maruadsi. The DPPH scavenging activity of the former was similar to the glutathione (Jiang et al., 2014). In addition, the CFCTKPC synthesized by Huang et al. (2012) contained three cysteines, and showed strong free radical scavenging ability and inhibited linoleic acid oxidation.

Histidine is one of the most common amino acids in antioxidant peptides. Its imidazole group can act as hydrogen donors to react with free radicals, especially superoxide anion-free radicals (Jie et al., 2019b). Kim et al. (2020) obtained histidine-containing low-molecular-weight (LMW) peptides from tuna waste meats. The histidine-containing LMW peptides exhibited a high DPPH radical scavenging effect in a dose-dependent manner.

Catalytic metal can accelerate the decomposition of oil oxidation primary products to generate free radicals as shown in Scheme 4 (Muhr et al., 2020).

Scheme 4. Transition metal ions accelerate the decomposition of oil oxidation primary products (Mⁿ⁺ represents transition metal ions).

The catalysis of catalytic metal depends on their ability to react with unoxidized oil to form hydroperoxides (Juita et al., 2011). For example, Fe²⁺ catalyzes hydrogen peroxide to generate hydroxyl radicals with high activity through the Fenton reaction, thus promoting oil oxidation (Walters et al., 2018). Another theory was that the presence of transition metal would significantly reduce the content of tocopherols, thereby affecting the oxidation stability of the oil (Fomuso et al., 2002).

Antioxidant peptides can reduce the chemical reactivity of metals by chelating with metal ions to form stable complexes. Antioxidant peptides can also block the interaction between metals and oil in space. For example, changing the physical position of metal ions to reduce the rate of free radical reactions indirectly achieves the antioxidant effect (Feng et al., 2016). Regarding antioxidant peptides, suitable amino acids and their correct position in the sequence play a crucial role in their chelating activity (Walters et al., 2018). The imidazole group of histidine in the antioxidant peptide sequence can chelate with metal ions, thus inhibiting the generation of free radicals catalyzed by them (Burkitt, 2001). Basic amino acids such as glutamic acid and aspartic acid (with carboxylic acid groups) can also inhibit the catalytic oxidation of transition metal ions (Sonklin et al., 2018). Besides, amino acids such as arginine and lysine with amino groups, cysteine with thiol groups and serine and threonine with hydroxyl may form complexes with antioxidant peptides by electrostatic interactions or H-bond coordination (Canabady-Rochelle et al., 2018; Walters et al., 2018). It has also been reported that casein-hydrolyzed peptides have the ability to inhibit the oxidation of oil. The peptides can oxidize Fe²⁺ into less reactive Fe³⁺ to delay oil oxidation (Diaz et al., 2003). Some antioxidant peptides with metal chelating properties can act as potential antioxidants because of their charged amino acids in their sequences. The charged amino acids can be electrostatically attracted to metal ions to inhibit their catalytic effect. For example, the positive charge of histidine can attract negative free radicals, which is consistent with the results of Saiga et al. (2013). It is worth noting that antioxidant peptides with molecular weight less than 3 kDa have less chelating force on Fe²⁺ than the peptides with the formula weight of 5-10 kDa (He et al., 2013). Therefore, antioxidant peptides with large molecular weight have a better effect on inhibiting oil oxidation.

Moreover, it must be noted that even though a peptide chelates metal, that does not mean it inhibits metal-promoted oil oxidation. Some chelators bind iron but do not decrease its reactivity (Durand et al., 2021). Therefore, additional assays should be conducted to better assess the ability of these peptides to inhibit oil oxidation induced by prooxidant metals.

Hydroperoxide is the primary product of oil oxidation. It can react with the ingredients in food, and reduce its quality. With the development of oil oxidation, hydroperoxide eventually decomposes into small molecules such as aldehydes and ketones (Chen et al., 2020a). Hydroperoxides may even decompose into toxic substances such as malonaldehyde (MDA) and 4-hydroxynonenal (HNE). During the chain transfer period, the free radicals generated by the decomposition of the hydroperoxides can react with unoxidized oil, continuously creating new free radicals and hydroperoxides to accelerate the oxidation of oil (Krugovov et al., 2014).

The acidic amino acids in the antioxidant peptide sequence form hydrogen bonds with unsaturated fatty oil to protect the double bonds of fatty acids and decrease the cleavage of C-H bonds, thereby reducing the generation of hydroperoxides (Jie et al., 2019b). A new antioxidant peptide was isolated from the Caragana ambiguas seed protein. The glutamine in the Caragana ambigua seed peptide (CSP) can form hydrogen bonds with unsaturated oil to protect the double bond of fatty acids, thus reducing the generation of hydroperoxides. In addition, CSP can delay the autoxidation of oil by protecting the phenolic compounds (Jie et al., 2019a). Antioxidant peptides can also reduce the reactivity of hydroperoxides by proton or electron transfer, thus inhibiting the chain transfer period of oil oxidation (Li and Yu, 2015; Lacou et al., 2016). Some researchers prepared an antioxidant peptide (APHPH) with the molecular weight of 1801 Da. The hydrophobic amino acids in its sequence can provide protons to inhibit the reaction between hydroperoxides and free radicals, and effectively inhibit oil oxidation in the linoleic acid emulsion system (Kim et al., 2007). An antioxidant peptide was prepared by using Alcalase alkaline protease and ultrafiltration technology from silkworm sericin. Silk sericin peptide has an inhibitory effect on the hydroperoxides induced by 4NQO (an oral carcinogen) in liver tissue (Fan et al., 2016). Other researchers believe that antioxidant peptides can also reduce hydroperoxides to relatively inactive hydroxides through non-radical reactions. The reaction mechanism is to generate sulfenic acid and sulfoxide derivatives through double electron transfer from the sulfur of cysteine or methionine (Esfandi et al., 2019).

In summary, antioxidant peptides can inhibit oil oxidation by scavenging free radicals, chelating metal ions that promote oxidation and reducing hydroperoxides generation. In addition to these common mechanisms, it has been proposed that antioxidant peptides may interact with secondary oxidation products of oil, such as aldehydes. These reactions may interfere with oil oxidation and alter the rate and pathway of oil rancidity (Berton-Carabin et al., 2014). However, since the secondary oxidation products of oil are bound to the antioxidant peptides and not available for determination (e.g., by GC headspace analysis), the mechanism remains to be further explored (McClements and Decker, 2018).