Covalent Flexible Peptide Docking in Drug Discovery: Current Challenges and Potential Interventions

Recently, there has been a significant increase in the search for new covalent inhibitors through drug discovery platforms, which has led to the development and implementation of new computational tools, including covalent docking methods to predict the binding mode and affinity of covalent ligands. Since the discovery of insulin nearly a century ago, more than 80 peptide medications have hit the market to treat a wide range of diseases, including diabetes, cancer, osteoporosis, multiple sclerosis, HIV infection, and chronic pain. Electrophilic peptides that form covalent bonds with target proteins have great potential for binding targets that have been previously considered undruggable. Despite the recent advancements in computational performance and docking algorithms, covalent docking still poses a number of challenges. Peptides covalent docking presents additional challenges, the main ones being the choice of the optimal peptide sequence, incorporation of electrophilic warheads, the inherent flexibility of peptide structures, and the fact that, unlike small molecules, peptides do fold. In this review, we present a brief overview of the current state of peptide therapeutics in drug discovery, covalent docking in general, covalent peptide binders, and-with particular emphasis-the difficulties that covalent peptide-protein docking is currently facing and some potential solutions.


INTRODUCTION
Even though covalent inhibitors have been used therapeutically for more than a century, there was widespread opposition in the pharmaceutical industry to their further development due to safety worries.After the creation of a wide range of covalent inhibitors for the treatment of human health conditions and the FDA's approval of several covalent therapeutics for use in humans, this disposition has recently been reversed [1] .In the past few years, there has been a significant increase in the search for new covalent inhibitors [1] .
Recent developments have also made it possible to use them as chemical probes to reveal new and harder-to-reach targets.The rising interest in covalent drug discovery prompted the development of new computational tools, such as covalent docking techniques, that are available to predict the binding mode and affinity of covalent ligands [1][2][3] .As a result of this expanding research into the discovery of covalent inhibitors, more than 80 covalent therapeutics have been approved by the FDA and currently about 30% of the marketed drugs are covalent inhibitors [1] .
Figure 1 displays some selected examples of covalent drugs that the FDA has approved and the year of approval.Acetylsalicylic acid, also known as aspirin, was the first covalent inhibitor to be commercialized in 1899, and it covalently binds to Ser-530 of prostaglandin endoperoxide synthase-1 via its acetyl warhead [2] .This was followed by the development of a vast array of covalent drugs for various disease.For instance, timoprazole demonstrated suppression of stomach acid secretion in 1975.However, further development of timoprazole was halted due to its toxicity [3] .Omeprazole, a derivative of timoprazole, was then identified in 1979 as the first medication class currently referred to as a proton pump inhibitor (PPI) [4] .Omeprazole covalently binds to hydrogen-potassium ATPase through disulphide bond.Lansoprazole, which was then introduced to the market in the early 1990s, and given FDA approval in 1995, hence was the second PPI medication to enter the market [2] .Two of the most popular covalent medications in the 1990s were clopidogrel and fosfomycin.Clopidogrel is an antiplatelet drug that attaches to the P2Y21 purinergic receptor irreversibly through the -SH group [5] .Through the inhibition of the enzyme MurA, fosfomycin suppresses the synthesis of bacterial cell wall.It is equipped with an epoxide warhead aimed against MurA's active site residue Cys115 [6] .One of the rare covalent inhibitors with a boronic acid warhead is the anticancer drug brentezomib.In 2003, it received approval and released to the market.Two of the most recently approved covalent inhibitors of BTK and EGFR are Zanubrutinib (approved November 2019, marketed under the trade name Brukinsa®) [7] and dacomitinib (approved September 2018, marketed under the name Vizimpro®) [8] .

Small Molecules Inhibitors vs. Peptides -Evolution of Covalent Peptides
Small molecules are typically used to target particular proteins, but in many instances, due to their size, flatness, and lack of natural substrates, some targets are notoriously challenging to drug using small molecules.Sites of protein-protein interactions, shallow allosteric pockets, transcription factors, and DNA binding proteins generally are examples of difficult protein targets [2] .Using peptides, which have a much larger surface area and can interact with "hot-spots" on the target's binding surface, is a common strategy for attacking such targets.Such peptides have a high degree of biological affinity and target specificity and safety [9][10][11] .1922 was the starting year of the field of peptide drugs, using animal insulin extracted from bovine and porcine pancreas to treat the type 1 diabetes [12] .Then, in 1954, the team led by Vincent Du Vigneaud introduced oxytocin as the first artificial polypeptide [13] .Bruce Merrifield took the next step, believing that peptide synthesis could be automated by using a solid phase to assemble different amino acids.This research led to the creation of Solid Phase Peptide Synthesis in 1963 [14] .
Although not usually in accordance with Lipinski's rule of five, recent chemical methodologies have been shown to enhance the pharmacokinetic properties of peptides, such as bioavailability, permeability, and in vivo stability.These chemical modifications may include N-methylation, cyclization, and incorporation of d-amino acids [15][16][17] .It is still difficult to find high-affinity peptide binders that can compete with natural cellular interactions and get around the inherent binding challenges of large protein surfaces [11] Longer chain peptides might be able to interact with more protein "hot spots", but doing so will increase the binding enthalpy and raise the entropic cost, which will have a significant detrimental effect on the binding affinity.Additionally, pharmacokinetic profiles for longer chain peptides are usually poor.Forming a covalent bonding between the peptide and its target is one possible approach that might be used to overcome this challenge, hence greatly increase the peptide's potency [18,19] However, compared to covalent small molecule inhibitors, the reported number of covalent peptides is still much lower.Although significant efforts have been made to aid the modelling and design of noncovalent peptides, the molecular simulations of covalent peptide interactions remain a challenging task, especially with the lack of peptide-specific validated protocol that takes into account covalent binding between the peptide and the receptor [18,[20][21][22] .Along with our earlier reports on covalent docking [23][24][25][26][27][28][29] , The current report will concentrate on covalent peptide docking, its challenges, and provide potential solutions including technical guidance to useful tools and approaches to assist with the rational design of covalent peptide inhibitors.

COVALENT VS. NON-COVALENT INHIBITION
In general, covalent inhibitors bind to the targeted protein and inactivate the protein temporarily or permanently [30] .In fact, the mechanism of the inhibition process depends on the reaction equilibrium between the ligand functional groups -that can be electrophilic or nucleophilic and the amino acids of the targeted protein [31] .As a result of this nature of interactions between the "warhead" and the protein active site, the complex can represent a bound or unbound states [32] , which results to one of two scenarios: covalent or noncovalent interactions [33] (Figure 2).
Interestingly, it was demonstrated that different drugs can bind with the same target via different mechanisms.For instance, pirtobrutinib, the first and the only noncovalent BTK inhibitor approved by FDA recently [34,35] , was found to bind non-covalently to BTK (Figure 3A), while zanubrutinib represents an example of covalent binder to the same protein, BTK (Figure 3B) [36] .

Covalent Inhibition Was Always Thought to Be "Irreversible" -Is This Still the Case?
Generally speaking, covalent inhibitors are compounds that, by design, are meant to form a covalent bond with a particular molecular target.It was always thought that covalent inhibitors are irreversible until the concept of

A B
covalent-reversible modification was recently introduced to the for several biological targets [37][38][39][40][41][42] .In order to maintain the desired maximum drug-target residence time on the target of interest, it is hoped that this modification to reversible nature will lessen the risk of unspecific covalent modification and, consequently, possibly lower the risk of severe side effects.
Depending on the selected warhead, the covalent bond may be either reversible or irreversible.Rauh et al have developed an efficient methodology to characterize the covalent and reversible characteristics of a library of covalent reversible inhibitors and investigated their potential in targeting tyrosine kinase EGFR [43] .The electron-withdrawing substituents in these compounds make the β-keto position more liable to nucleophilic attack, hence quickening the adding process in order to bind the target protein covalently.Furthermore, these substituents make the Cα-H in the covalently attached inhibitor more acidic, which makes it easier to quickly remove the proton when the protein-inhibitor environment changes [44,45] .Figure 4 illustrates an example of reversible covalent inhibitor HKI-272 in a complex with EGFR kinase domain T790M/L858R mutant [46] .

COVALENT DOCKING OF PEPTIDES IN DRUG DISCOVERY -NOT TRIVIAL!
We have previously discussed covalent docking in detail, including its uses and applications, available tools, advantages and disadvantages [25][26][27] .To avoid repetition of information, the primary objective of this review is the covalent docking technique specifically related to peptidebased molecules.But first, a number of important aspects must be thoroughly grasped before discussing covalent docking approach for peptides.These include the choice of amino acid sequence, the structural and chemical prerequisites for a peptide to create a covalent link with a target protein, flexibility and so on.In the following text, we discuss each of these aspects, which all together would help build up a sense of understanding of rational design of peptide-based covalent inhibitors.

Peptide Structural Requirements -Peptide Sequence, Electrophilic Warhead Flexibility and Folding Challenges 3.1.1 Rational Design of Peptide Sequence
The physicochemical and ADMET qualities of a

A B
peptide are determined by the amino acid sequence that forms it.The physicochemical characteristics of peptide-based medications can be tailored using a variety of techniques to overcome issues such as low permeability, metabolic instability, and limited tissue residence duration [47] .In this context, the use of bioinformatics, artificial intelligence, and computational modeling applications has been crucial 48] .The variety and order of amino acids can be intelligently chosen based on the research goal.Here we present some useful tools that help with the prediction of peptide sequence, physicochemical and toxicity profiles of the designed peptides.Peptide combination generator server (https:// pepcogen.bicfri.in/) is one useful tool that can be used to create any combination that a specific peptide sequence based on its physiochemical characteristics.ToxinPred (https://webs.iiitd.edu.in/raghava/toxinpred/index.html) is an in silico method, which is developed to predict and design toxic/non-toxic peptides.Database of Antimicrobial Activity and Structure of Peptides (https://dbaasp.org/) is the database that is manually curated.It is created to give scientists the knowledge and analytical tools they need to specifically create antimicrobial compounds with a high therapeutic index.PEP-FOLD (https://bioserv.rpbs.univ-paris-diderot.fr/services/PEP-FOLD/) is a de novo approach aimed at predicting peptide structures from amino acid sequences.This approach combines the projected series of SA letters to a greedy algorithm and a coarse-grained force field, based on the structural alphabet SA letters to characterize the conformations of four successive residues.PrMFTP (http://bioinfo.ahu.edu.cn/PrMFTP/) is another a web application used to identify multi-functionalities of therapeutic Peptides.More precisely, PrMFTP learns and fully extracts informative features from peptide sequences in order to anticipate multifunctional therapeutic peptides by utilizing multiscale convolutional neural networks, bi-directional long short-term memory, and multi-head self-attention techniques.

Electrophilic Reactive Warhead
A warhead refers to the chemical moiety that forms a covalent bond with the target protein through the classical electrophilic-neutrophilic reaction with nucleophilic amino acids in the protein binding site such as cysteine, lysine etc. Selecting the right warhead or modification is essential since it maximizes the inhibitor's effectiveness, reactivity, and selectivity with the target protein while reducing its off-target effects [49][50][51] .In contrast to small molecules, which might naturally have electrophilic groups in their chemical structures, peptides lack an electrophilic moiety by nature, hence it is necessary to add one in order for a peptide to form a covalent connection with a nucleophilic amino acid, most commonly cysteine and lysine, in the binding site of target protein.Common examples of warheads include acrylamides, thioesters, alkynylbenzaldehydes, sulfanamides, and boronate-based reagents, among others are shown in Figure 5.
Determining the precise chemical groups that will interact with the target protein is necessary for choosing the reactive electrophilic moiety in peptide design [52] .Example of chemical modification of peptides to include electrophilic warhead is shown in Figure 6.
In this regard, CovPepDock (https://rosie.graylab.jhu.edu/cov_pep_dock) is a useful predictive tool that can be used to identify an optimal warhead and create peptide covalent modifiers (warheads) by using a known noncovalent binder as a starting point to make irreversible covalent link with a target cysteine [54] .

Flexibility Challenge
Unlike small molecules, besides their conformational flexibility, peptides often have conformational preferences.Many short peptides in solution can adapt their native secondary structures, α helices, and β hairpins [55] .The fact that peptides have inherent conformational flexibility must therefore be taken into account when performing simulation studies.Several computational docking algorithms have been developed to address these challenges associated with peptides simulations.These can be split into to three classes (Figure 7): templatebased docking, local docking and global docking [56] ..

Template-based Docking
The template-based docking is based on the exploitation of known templates -scaffolding proteins-to predict the model protein-peptide complex [21,57,58] .This docking category could be carried out by using a range of tools to analyse sequence-structures manually, semiautomatically or fully automatically.If the investigated complex and the used template are quite similar, this category may be appropriate, successful, and useful [56] .One tool that applies this protocol is GalaxyPepDock webserver which predicts protein-peptide interactions based on templates [57] .It uses protein structure in standard PDB format and a peptide sequence as FASTA format as input files and predicts as output the proteinpeptide complex structures [57] .
The GalaxyPepDock runs template-based peptideprotein docking in two separate phases [21] .Primarily, the selection of templates by searching in the PepBind Database or PDB which contain structural information for peptide-protein complexes [59] , the choosing stage is based on similarities of two scores, the structure of the protein of interest measured by the template modelling score "TMscore" [60,61] , and the peptide-protein interactions measured by interactions similarity score "S inter ".The top ten complex scores are selected as templates if the score is higher than 90%, those selected templates are used for model building.Secondly, the complex structure optimization, for each selected template, 50 models are generated by the model-building tool of the Galaxy template-based modelling based on the GALAXY molecular modelling package [62,63] .Then, only 10 structures are chosen by the GALAXY energy, and they are either exposed or refined by the GalaxyRefine module [64,65] .

Local Docking
Local docking techniques run docking close to the protein of interest's known binding active site, however the accuracy of the results depends heavily on the input data.A numerous programs have been developed that require less strictly defined input model, such as Rosetta FlexPepDock.It is one of the first protocols to explicitly include peptide flexibility in docking [66] .DynaDock is another web server for peptide-protein docking that also include receptor flexibility into account [67] .PepCrawler is another tool for high binding affinity of peptide inhibitors that adopts local docking algorithm [68] .

Global Docking
On the other hand, using the first or second class of computational protein-peptide docking methods to predict complex structures is quite difficult when information on the nature of the binding interactions is not available [69] .In such cases, global docking methods are used in.To dock peptide molecules, this class searches the entire protein surface, so a predefined binding site or peptideprotein complex structure are not required inputs [57] .This search pipeline typically consists of three phases: the prediction of input peptide conformations, docking and model generation, and scoring the models and refinement phase.However, other methods may be used.PEP-SiteFinder [70] and CABS-dock [71][72][73] are two examples of web servers used for global docking that do not require information about the binding site.

Flexible Peptide, Flexible Protein -A Major Challenge in Docking in General
Structural flexibility of the ligand (a peptide in this case) and target protein is considered the most challenging aspect facing docking simulation at several  [56] .levels [74] .Protein-peptide complexes are usually formed by involving the folding of the flexible peptide meanwhile binding to the targeted protein [75] .A higher percentage of flexible bonds makes docking operations more difficult and produces unreliable results.Indeed, the number of peptide's flexible bonds affects directly the prediction accuracy.Taking into account the dynamic nature of proteins, receptor flexibility is identified as an important factor for peptide-protein interactions [58,76] .In fact, during the binding event, protein flexibility may change from side chain bonds to large scale backbone [77] .Numerous methods ignore the receptor's flexibility in order to reduce computational costs, instead using rigid-body protein docking [56] .For example, CrankPep [78] adopts this approach for the docking of flexible peptides to rigidbody receptors.However, taking receptor flexibility into account directly affects the docking process, leading to inconsistent docking results [79][80][81] .
Numerous flexible peptide-protein docking protocols have been created to address this challenge, and in the text that follows, we highlight the various tools available for flexible peptide-protein docking.Reference [88] HPEPDOCK allows for partially side chain flexibility -with a possibility to include the protein backbone -such as PepCrawler [68] , DINC 2.0 [82] , AutoDock Vina [83][84][85] , Rosetta FlexPepDock [86] Fragments-based docking implemented in PIPER-FlexPepDock [87] , is one useful tool for flexible peptideprotein docking that applies global docking algorithm with a high resolution approach modelling of high peptideprotein interactions leading to detailed analyses of theses interactions by using fragment ensembles [87] .To consider the receptor flexibility, there is other class of protocols using soft-core potentials imitators of implicit receptor flexibility, including DynaDock [88] .In addition, CABS-Dock server is used for flexible peptide-protein docking, but it is limited to small backbone fluctuations [71][72][73] .HADDOCK [89] is one of the tools used for flexible proteinpeptide docking. 2 considers the peptide flexibility through an ensemble of peptide conformations generated by our MODPEP program [90] .
There are several protocols designed specifically to mimic covalent peptide bonding.One widely used tool is Rosetta CovPepDock which is a high-resolution peptideprotein docking refinement protocol for the modelling of peptide-protein complexes that was developed specifically for covalent flexible peptide [91] .Molsoft incorporates modules for covalent flexible peptides docking [92] , Schrodinger CovDock [93] , Gold [94] , DOCKTITE-MOE [95] and AutoDock4 [96] , and FITTED [97] are also commonly used for covalent docking, along with other tools that are summarized in Table 1.Comparative evaluation of covalent docking tools is demonstrated in this report [98] .

Sampling and Scoring Function-The Trade-off Between Accuracy and Computational Cost
Sampling and scoring functions are two pillars of each existing docking protocol, However, they are directly related to numerous challenges as they influence the docking result accuracy by a combined effect [76] .Sampling is a crucial docking step because drug discovery requires attention to sampling, speed, and acceleration.There are four major classes of sampling methods classified based on the level of receptor flexibility: (1) The simplest category, soft docking, treats implicitly only a portion of target flexibility (cannot be used for large side chain or backbone); (2) selective docking is based on choosing only a few crucial residues that are linked to the peptide receptors.This approach might work if the receptor's structural information is available; (3) ensemble docking usually considers full flexibility of the receptor implicitly; (4) on-the-fly docking considers receptor flexibility explicitly as it generates new receptor conformations on the fly during the docking process, however, treating receptor and peptide flexibility simultaneously increase the complexity of the operation.Detailed information on sampling techniques can be found in this source [76] .
On the other hand, the scoring functions also contributes to the accuracy of docking results.scoring function is a ranking system consists of binding interaction estimation (binding energy) for a specific bond [99] .In fact, for flexible protein-peptide docking methods, scoring function takes the position of every peptide atom and each protein atom and converts it to a numeric value as a score, this is the procedure to rank the potential conformations of docking results, toward to select the most reasonable conformations to symbolize flexible peptide-protein complex in vivo [100] .However, to identify the highest accurate complex from all docking results is the challenge in this operation [56] .
In addition, the crucial difficulty facing docking approaches with covalent peptides is the formation of covalent bonds between the electrophile peptide's functional groups and the nucleophilic residue in the active site of a receptor.The biggest portion of docking programs specific to covalent docking are incapable to perform an accurate estimation of covalent binding energy, because the covalent bond formation energy is excluded from the general score.In general, three classes of scoring methods can be distinguished: (1) force-fields based scoring functions; (2) empirical or semi empirical scoring functions; (3) knowledge-based functions that are elaborated by using statistical data of well-known structures of peptide-protein complex [101] .CAPRI is communitywide experiment on the comparative evaluation of protein-protein docking for structure prediction demonstrated that the best peptide-protein docking results are generated by using methods established on mixed scoring functions [58,102] .
Finally, integrating experimental data into computational calculations has been implemented in various docking scoring functions and algorithms in order to improve docking accuracy.A number of docking programs have included cryo-EM data into their operational procedures.For example MultiFit automatically segments the cryo-EM density using a Gaussian mixture model to deduce anchors, subsequently docking the components of the complex onto the anchors [103] .Using 3D Zernike descriptors, EMLZerD scores the models based on the cryo-EM data [104] .A recent method has been used in ATTRACT-EM that fits the subunits into the map using a Gaussian mixture model to represent the cryo-EM data, and then refines the resulting models [100] .However, it is challenging to convert the experimental data into computational information of the covalent peptide-protein interactions during the docking process.Apparently, these approach encounters a lot of difficulties and ambiguities in terms of the conversion robustness of the available experimental data [101] .Most of these methods separate the use of the cryo-EM data from the use of other sources of information.Some important aspects such as physico-chemical properties (energetics) are only taken into consideration after fitting or refining structural parameters.Additionally, they typically do not make active use of any additional orthogonal data that might be provided, like mass spectrometry crosslink data or mutagenesis data.HADDOCK is one useful tool that integrates Cryo-EM and other experimental data in its docking algorithm to enhance docking scoring functionality [89,105] .

Flexible Protein-peptide Docking Methods -A Way Forward
Simulations of covalent inhibitor binding remain a challenging topic, with many technical considerations to be taken into account.The starting point of performing covalent binding simulations is covalent docking.Although molecular docking methods in general have undergone significant development over time, covalent flexible docking, especially for peptides, remains a huge technical challenge.Docking accuracy depends on many factors, such as the size and architecture of peptideprotein complex, the, sampling method, the choice of parameters and scoring functions and the flexibility of both the protein and peptide.Peptides conformational flexibility and folding motif present another worrying challenge.Besides, there is no unique protocol or specific criteria that could be possibly used to obtain the best results.Another obstacle might be the protein lowresolution crystal structure or a missing amino acid sequence.Lack of information on binding sites adds another difficulty to the process.Overall, the accuracy of predicting a peptide-protein complexes is the main object, and the major difficulty for any predictor is modelling the bound conformation of the protein-peptide interactions with best accuracy levels.
We believe that a concise technical guide on each challenge and a list of potential solutions would be helpful to researchers working on improving docking protocols (Table 2).These guidelines are primarily based on our experience we have accumulated over the years in covalent docking research [23][24][25][26][27][28][29] .
All in all, it is expected that the forthcoming protocols would be able to combine more than one approach, providing facilities to integrate experimental and computational docking results.The advances in computational performance and the integration of Artificial Intelligence and Machine learning would improve Homology modelling approach can be used to address this challenge.A few technical tips on homology modelling of protein structure are provided in refs [106,107] .AlphaFold Artificial Intelligence (AlphaFold AI) is another useful tool for prediction of protein 3D structures [108] .

Missing residues in the protein structure
There are several tools to address this challenge.SL2 is one of the easy-to-use and efficient tools that can be used.It is a fragment-based tool for the prediction and interactive placement of loop structures into globular and helical membrane proteins [109] .
No available information on the binding site Binding site prediction tools can be used.A wide range of tools that can be used for this purpose and validations protocols are available.See provided references [110,111] .

Flexibility/conformations of peptides
It is essential to determine the lowest energy conformations of peptides and most importantly their folding motif [112] .A few techniques have been developed for modelling protein-bound peptide conformations, Here we present a list of tools that can be used to predict peptide structures and folding motif: MODPEP can be used for conformational sampling of proteinbound peptides [113] .PEP-FOLD (https://bioserv.rpbs.univ-paris-diderot.fr/services/PEP-FOLD/), APPTEST (https://research.timmons.eu/apptest_help)and PEPstrMOD (https://webs.iiitd.edu.in/raghava/pepstrmod/) are useful tools for predication peptide structures.Two review articles tackle problems related to peptides such as predicting structures, binding affinities and even kinetics [112,114] would also serve as useful resources in this regard.
Performing molecular dynamics simulations of covalent bonding is technically challenging, however, we recently published a detailed technical guide on how to perform all-molecularmechanics MD for covalent simulations [115] .

Flexible peptide, flexible protein
Although there are several docking tools that take flexibility of both protein and peptide into account (listed in Table 1), it is strongly recommended that molecular dynamics simulations can be used in conjunction with docking calculations to verify docking results.A few published protocols can be accessed here [116][117][118] .

Too many scoring functions and algorithms-which one to pick?
There is no typical protocol/algorithm that can guarantee the accuracy of docking calculations, hence it is essential to validate docking results using one or more approaches described below.Some useful reports on the covalent docking protocol selection are provided [98,119,120] .

Validation of docking protocola must to do
There is no doubt that validation of docking results against experimental data is the most acceptable approach, however, several in silico validation protocols can be adopted: Cross-validation by using more than one covalent docking tool to cross-validate the results.Re-docking of experimentally determined X-ray crystal structure of peptide-protein complexes to assess the accuracy of docking in predicting the native binding pose.However, careful attention should be paid here as re-docking can be an artificial exercise.Incorporation of molecular dynamics simulations and thermodynamic calculations such as MM-PBSA/MM-GBSA [121] to estimate the binding affinities.
Ensemble docking is a method that is used to generate an "ensemble" of drug target conformations, frequently through the use of molecular dynamics simulation.
the accuracy of docking calculations.The molecular docking process, however, is a multi-dimensional one that calls for a logical protocol, design, and validations-most importantly, against experimental data or counterparts.

CONCLUSION
Electrophilic peptides that can attach irreversibly through a covalent bond with target protein offer considerable potential for drug discovery especially for protein targets that were previously thought to be undruggable.Nevertheless, the discovery of such covalent peptides still poses several challenge, hence efficient de novo rational design approaches for covalent-based peptide inhibitors are critical.In this context, herein, we offered a "one stop shop" for in silico rational design of covalent-based inhibitors, associated challenges and possible interventions and expert opinion.This review discussed covalent docking of peptide-based molecules from a technical point of view, as well as a number of theoretical facets that, all together, provide a broad yet crucial knowledge of the principles of rational design of peptide-based inhibitors.The various aspects of peptide covalent docking and flexible peptide-protein docking were also highlighted in this review.Determination of the optimal amino acid sequence serves as the initial yet most crucial starting point of design of peptide-based inhibitors.The peptide sequence can be proposed based on chemical intuition, experimental data, existing peptide drug template, quantitative structure activity relationship, the binding site architecture of target protein, proteinprotein interface, etc.It is also crucial to consider that the amino acid sequence of a peptide determines its physicochemical characteristics, toxicity profile, pharmacodynamic landscape, and capacity to bind to a protein target.Because they don't have electrophilic moieties, natural peptides can't bind covalently to target proteins, therefore, electrophilic warhead must be incorporated in the structure of the peptide for covalent binding to occur.For this, we showcased the available in silico and bioinformatics tools to assist with the de novo design of peptide sequence, selecting electrophilic warhead and the most reliable tools for covalent docking, taking into account peptide flexibility and folding.We also presented the major challenges of flexible peptide covalent docking and provide potential practical solutions and intervention based on our experience in the field.
We believe that this report adds significantly to the understanding of rational design of peptidebased covalent inhibitors and their applications in drug development and discovery.

Figure 1 .
Figure 1.Some examples of the FDA-approved covalent drugs and the year in which it was approved.

Figure 5 .
Figure 5. Common examples of warhead chemical structures that are used as peptide modifiers: (A) cysteinereacting and (B) lysine-reacting.The dotted line in the structures above is the attachment point to peptide chain.

Figure 6 .
Figure 6.An example of chemical modifications of peptides to incorporate an electrophilic warhead.The 2D chemical structures were taken from[53] .

Figure 7 .
Figure 7. Classification of protein-peptide docking methods and some examples of the tools used for peptideprotein docking[56] .

Table 1 . Different Tools Used for Flexible Non-Covalent and Covalent Docking and Their Access Links Peptide-Protein Docking Tool Overview Link Non-Covalent Peptide-Protein Docking
Table 1 includes a list of these techniques/tools in summary form.Some tools