Skip to main content
NCBI home page
ACS AuthorChoice logoLink to ACS AuthorChoice
. 2022 Dec 27;63(1):187–196. doi: 10.1021/acs.jcim.2c01261

pIChemiSt — Free Tool for the Calculation of Isoelectric Points of Modified Peptides

Andrey I Frolov †,*, Sunay V Chankeshwara †, Zeyed Abdulkarim ‡, Gian Marco Ghiandoni §
PMCID: PMC9832473  PMID: 36573842

Abstract

graphic file with name ci2c01261_0008.jpg

The isoelectric point (pI) is a fundamental physicochemical property of peptides and proteins. It is widely used to steer design away from low solubility and aggregation and guide peptide separation and purification. Experimental measurements of pI can be replaced by calculations knowing the ionizable groups of peptides and their corresponding pKa values. Different pKa sets are published in the literature for natural amino acids, however, they are insufficient to describe synthetically modified peptides, complex peptides of natural origin, and peptides conjugated with structures of other modalities. Noncanonical modifications (nCAAs) are ignored in the conventional sequence-based pI calculations, therefore producing large errors in their pI predictions. In this work, we describe a pI calculation method that uses the chemical structure as an input, automatically identifies ionizable groups of nCAAs and other fragments, and performs pKa predictions for them. The method is validated on a curated set of experimental measures on 29 modified and 119093 natural peptides, providing an improvement of R2 from 0.74 to 0.95 and 0.96 against the conventional sequence-based approach for modified peptides for the two studied pKa prediction tools, ACDlabs and pKaMatcher, correspondingly. The method is available in the form of an open source Python library at https://github.com/AstraZeneca/peptide-tools, which can be integrated into other proprietary and free software packages. We anticipate that the pI calculation tool may facilitate optimization and purification activities across various application domains of peptides, including the development of biopharmaceuticals.

Introduction

Peptides as Therapeutics

The number of FDA-approved peptide therapeutics (PTs) has been steadily growing over the past decade.1 Peptides and peptidomimetics fill an intermediate niche between conventional small molecules and large biologic therapeutics.2 Compared to small molecules, PTs allow targeting of shallow pockets and are usually less promiscuous. Compared to biologics, they are less immunogenic and have lower production costs.

Chemically Modified Peptides Are Widely Spread in Nature and Industry

Bioactive peptides originate from various natural sources (e.g., endogenous hormones,3 toxins,4 etc.), hit-finding approaches (e.g., phage display),5,6 and rational design (e.g., epitope mimetics).7,8 Peptide starting points are often modified with nCAAs9 to modulate the interaction with the target surface10 and improve in vivo stability and oral bioavailability.11−13 Short cross-linked peptides mimicking interaction epitopes are often designed for the inhibition of protein–protein interactions (PPIs).7,8,10,14,15 Peptides chemically linked with a cargo molecule can serve as targeted delivery vehicles to the tissues of interest.16 In addition to these applications, medicinal chemists are also constantly searching for noncanonical peptide mimetics and peptide-hybrids, such as peptoids,17 peptidomimetics,18 foldamers,19 and peptide-natural product combinations for their use as therapeutics.20 In turn, living organisms provide an immense source of toxins, venoms, and other bioactive peptides that are often highly modified.4 Additionally, nature provides the whole machinery of chemical modification of canonical amino acids with various post-translational modifications (PTMs).21,22

Challenges with Peptide Purification, Separation, and Immunogenicity

Handling peptides with natural and noncanonical amino acids in the lab faces practical challenges: their synthesis often requires multiple steps resulting in impurities that are structurally very similar,23 making purification difficult and time-consuming.24,25 Reverse-phase HPLC is widely employed in peptide purification but is burdened by its own challenges related to strong retention of hydrophobic peptides, sustainability, and cost, particularly on a large scale.26 Conversely, ion-exchange (IEX) chromatography can be used to purify peptides with similar lipophilicity but differing in ionization states using appropriate mobile phase conditions.27−31 Note that peptide ionization has a significant effect on solubility and aggregation at different pH32 that may result in undesired immunogenicity and toxicity.33−36

Isoelectric Point to Address the Challenges

Issues related to the physical properties of peptides can often be addressed by analyzing their charge states. First, keeping the pH of the mobile phase away from the isoelectric point, for example, by adding acidic ion-pairing agents,37 reduces undesired lipophilic interaction with the column during purification.24,32 Second, in some cases, impurities, even being structurally similar to the desired product, differ by their isoelectric point or, more generally, by their net molecular charge at different pH. By analyzing synthesis and mass spectrometric data, it is possible to make a qualified guess as to which amino acid could have failed to couple or coupled unprompted, that is, missing/added in the sequence.38−40 A wise selection of the pH of the mobile phase, based on the difference in calculated pI of the target peptide and the anticipated byproducts, allows the separation of chromatographic peaks even for structurally similar compounds.24,31 Overall, one can improve peptide solubility and physical stability of peptide solutions by designing the isoelectric point away from the relevant pH range.32 In addition, predicting ionization states are often attempted to rationalize peculiar observations in biological and biophysical assays. For instance, Zapadka et al. associated a highly unusual pH-induced switch in GLP-1 aggregation kinetics to the protonation/deprotonation of the N-terminus.33

Lack of Tools Predicting Isoelectric Points of Modified Peptides

Despite the evident need for pI prediction methods for modified peptides, there is still a lack of free and robust tools. There are tools that provide isoelectric point predictions for natural peptides in the form of web services,41−43 free or licensed standalone software.44−47 A few more tools support some common PTMs but only in a limited number.48,49 In addition, Bjerrum et al. published a tool that can handle noncanonical amino acids.46 Its limitations, however, are that the method relies on licensed software50 for structure-to-sequence conversion and only implements a small set of 17 rules to predict pKa of noncanonical side chains. Moreover, the lack of a front-end service to facilitate its usage further hinders its uptake.

pIChemiSt—New Method Predicting Isoelectric Points of Modified Peptides

In the present work, we describe the development of an open source tool called pIChemiSt for isoelectric point calculation of natural and modified peptides. The tool integrates two programs for the prediction of ionization constants of nCAAs, namely, the licensed ACDlabs51 and pKaMatcher developed in-house to avoid dependencies on proprietary software. The latter is validated to cover a set of around 300 noncanonical amino acids available for peptide synthesis in the AstraZeneca chemistry stock. We also describe the curation of a data set of experimental isoelectric point values for modified peptides from the Reaxys database.52 Finally, we demonstrate that the predictions from our tool provide substantially superior accuracy compared to the plain sequence-based approach. The tool is released free of charge on GitHub.53 In addition, an associated web server is under development and will be described in a future publication.54

Methods

Calculating Charge versus pH Curves

The linear combination of the Henderson–Hasselbalch equations is commonly used for the calculation of protein or peptide charges as a function of pH (eq 1). The equation requires a set of apparent pKa values and types of dissociation (acid or base) of all ionizable groups in the molecule

graphic file with name ci2c01261_m001.jpg 1

where the two sums run over all basic and acidic groups of the molecule, correspondingly; pKai is negated logged acid-ionization constant of the acid or the conjugated acid of the base; Qconstant represents the permanent charge of the molecular entity associated with the constantly ionized molecular groups unless the molecule undergoes chemical decomposition, such as quaternary ammonium and alkyl pyridinium cations. The charge versus pH curves are illustrated in Figure 1.

Figure 1.

Figure 1

Open in a new tab

(a) Charge versus pH curves resulting from the Henderson–Hasselbalch equations for the dipeptide shown in (b). Different curves correspond to various predefined pKa sets of canonical amino acids. The magenta dots (pKa (acid1), pKa (base1), pKa (base2)) mark the inflection points, which correspond approximately to the pKa values of the ionizable groups for the case of the IPC_peptide pKa set. The white dot corresponds to the isoelectric point of the molecule. The isoelectric interval denotes the span of pH within which the predicted peptide charge is negligible. This parameter may be used instead of the calculated pI when the latter is poorly defined. The variation between different pKa sets allows judging the uncertainty of the calculated pI, which is reported as mean value ± standard deviation in the output. The group identified as ‘base2’ belongs to an N-terminus of a noncanonical amino acid; therefore, its pKa does not belong to the predefined pKa sets and is calculated on the fly. (b) Chemical structure of a peptide with three ionizable centers, C-terminal acid, basic N-terminal amine, and the histidine side chain.

Calculation of Isoelectric Point

The isoelectric point, by definition, is the pH at which the net molecular charge is zero. Therefore, pI can be found by solving eq 2 numerically, where Q(pH) represents the net molecular charge

graphic file with name ci2c01261_m002.jpg 2

In this work, we use the bisection method with an initial pH interval that goes from 0 to 14 and a charge tolerance of 0.01e. There are examples where the Q(pH) curve does not intersect the zero line, for example, when there are no ionizable centers in the molecule or there are either only basic or acidic residues in the sequence. In these cases, pI is technically not defined, and hence, no value is returned. We also capture the cases when the curve approaches the zero line from one side asymptotically. In these cases, we set a threshold of 0.05e as an indicator that the molecule “becomes uncharged”. Another case where pI is poorly defined is when there is a large separation between pKa of basic and acidic groups in the peptide. In this case, the Q(pH) curve crosses the zero line, but the slope of the curve is negligible, making it challenging to pick a pI value from a large pH interval. To characterize such cases, we introduce the concept of “isoelectric interval”, which is defined as a pH span where the net charge of the molecule can be considered negligible. Hence, we set the following interval for the charge being between −0.05 and 0.05e.

Determination of Ionizable Groups in the Molecule

The input structure is processed as described in Figure 2. First, we cut all peptide bonds and cap the resulting N-termini with acetyls and C-termini with methyl ketone groups. The capping allows modeling the effect of peptide backbone on predicted pKa values of side chains, which might be significant as the carbonyl groups have strong electron-withdrawing character. Second, we match the canonical amino-acid side chains using a built-in set of SMARTS patterns (see Table S1 in SI). Note that the SMARTS patterns list hydrogens, tautomeric and ionization forms of each natural amino acid explicitly. Ionizable N-terminus amines and conventional C-termini (carboxylic acid and primary amide) are also encoded in the SMARTS patterns. The corresponding pKa values of ionizable side chains and termini are taken from the predefined sets (see the following section). Third, for the fragments that have no matches against the SMARTS patterns, we run automated pKa predictions (see the following sections). Finally, we merge the pKa sets from natural amino acids and unmatched fragments and use them to calculate charge versus pH curves and pI. Note that individual amino acids, where both termini are ionized, are not encoded in the SMARTS patterns, and thus, they are treated as noncanonical fragments by the software.

Figure 2.

Figure 2

Open in a new tab

Workflow of the pI calculation. The input chemical structures are analyzed, and all secondary and tertiary amide bonds are cleaved. The resulting amines and aldehydes are capped with acetyl and methyl groups, respectively. The resulting fragments are matched with the templates of canonical amino acids. The corresponding pKa values of the side chain and N- or C- termini are attributed from a look-up table. When a fragment is not recognized as a canonical amino acid, the method predicts ionizable centers and their corresponding pKa values. The pKa values of noncanonical fragments and canonical amino acids are combined and inserted into eqs 1 and 2 to obtain the charge versus pH curve, isoelectric point, and interval.

pKa Sets of Natural Amino Acids

There are many pKa data sets reported for natural amino acids, either derived from experimental measurements (Grimsley,55 Thurlkill,56 Toseland57), trained to reproduce experimental data (IPC_Peptide, IPC2_Peptide, IPC_Protein, and IPC2_Protein42,43), or obtained by combining experimental data with Hammett–Taft-derived electronic effects (initial Bjellqvist,41,58 extended Bjellqvist,59 Gauci60). The nature of other sets commonly used as benchmarks could not be deduced from the literature (ProMoST,48 DTASelect,61 Rodwell,62 EMBOSS,63 Nozaki64). The pKa values from the textbooks most likely refer to the experimental data of individual amino acids derived by potentiometric titration (Lehninger,65 Solomons66). Most pKa sets contain nine pKa values, two for N- and C-termini and seven for ionizable amino-acid side chains. There are also sets where the position of amino acids within a sequence influences their pKa values: Bjellqvist et al. introduced a 17-parameter set where the pKa values of the N- and C-termini vary depending on the type of the termini side chains.59 Later, Gauci et al.60 introduced two empirical values to the extended Bjellqvist set59 as they observed a systematic deviation for predictions when the Cysteine or Asparagine were at the N-terminus. Note that the pKa values in the Gauci set are not provided in their original publication; however, they are included in the source code of the pIR software.44,45 Finally, ProMoST48 uses a more advanced algorithm where the pKa values of side chains vary between C- and N-termini and the rest of the sequence results in around 56 pKa values. The origin of the pKa values implemented in ProMoST is not disclosed in its publication.48 The sets implemented in this work are listed in Table 1.

Table 1. pKa Sets Used in the Study.

  N-term. C-term. C Y D E H K R origin
IPC_peptide 9.564 2.383 8.297 10.071 3.887 4.317 6.018 10.517 12.503 machine learning
IPC2_peptide43 7.947 2.977 9.439 9.153 3.969 4.507 6.439 8.165 11.493 machine learning
Gauci44,60a 6.50–8.36 3.55–4.75 9.0 10.0 4.05 4.45 5.98 10.0 12.0 isoelectric focusing, Taft calculations, literature data
Grimsley55 7.7 3.3 6.8 10.3 3.5 4.2 6.6 10.5 12.04 experimental data (primarily NMR) of folded proteins
Toseland57 8.71 3.19 6.87 9.61 3.6 4.29 6.33 10.45 12.0 experimental pKa data (primarily NMR) of folded proteins
Thurlkill56 8.0 3.67 8.55 9.84 3.67 4.25 6.54 10.4 12.0 potentiometric titration of pentapeptides
Lehninger65b 9.69 2.34 8.33 10.0 3.86 4.25 6.0 10.5 12.4 potentiometric titration of amino acids
ProMoST48c 6.67–8.36 3.17–3.98 8.00–9.00 9.34–10.34 3.57–4.57 4.15–4.75 4.89–6.89 9.8–10.30 11.5–12.5 not disclosed
Open in a new tab
a

The Gauci set is based on the extended Bjellqvist set59 with two additional pKa values for N-terminal Cys and Asn. The pKa values in the extended Bjellqvist set were obtained from the experimental measures of the pI of the human carbonic anhydrase enzyme. Ionization constants of C-terminus, Asp and Glu side chains, were calculated using the Taft equation. The pKa of the N-terminus was derived from literature. No data were reported for the remaining amino acids as their pKa values fell outside the pH ranges considered in the study.58 However, these were listed in the subsequent publication, where they were derived from other literature sources.59 The pKa of the N- and C- termini in this set depends on the type of amino acid.

b

The pKa values of the side chain and N- and C-termini are reported for each amino acid. The origin of the data is unclear. However, it may result from potentiometric titration. The values for the N-terminus and C-terminus are averaged between all amino acids.

c

The pKa of the N- and C- termini depends on the type of amino acid. The pKa values of the side chains depend on the location of amino acids, where values at N- and C-termini differ from the rest of the sequence. We provide the span of values in the table. The pKa values can be found in publicly available data sources on the web.53,67,68

Prediction of pKa for Noncanonical Amino Acids

When a fragment is not recognized as a canonical amino acid, we analyze whether it has ionizable or constantly charged groups and predict the pKa of each ionizable center. Here, we implemented two options for such a purpose: the licensed ACDlabs51 GALAS method and an in-house open source tool called pKaMatcher, which can be used as an alternative to ACDlabs. Sometimes predictions include pKa of groups that change their ionization only at extreme pH ranges, which fall outside those typically evaluated experimentally (pH 2–12).69,70 We deliberately excluded these extreme pKa values from calculations as follows: pKa of acids should be below 12 and pKa of bases should be above 2 to be included in the calculation. Additionally, we set the lower limit for the acidic pKa to be −5 and the upper limit for the basic pKa to be 15. These limits are meant to keep strong relevant acids and bases in the calculation, for example, sulfate or guanidinium groups.

Prediction of pKa with ACDlabs

The tool executes the Percepta Batch (perceptabat) module of ACDlabs.51 We observed a misassignment of the type of pKa equilibria (acid versus base) for one of the non-natural amino acids while testing the CLASSIC algorithm, which is based on the Hammett–Taft equations. Therefore, we adopted the GALAS algorithm, despite being less documented. From the perceptabat output, we read all the predicted pKa values and the corresponding type of equilibria (acid or base) needed in eq 1.

Prediction of pKa with pKaMatcher

We developed a simple SMARTS pattern-matching approach for assigning pKa values of noncanonical amino acids and unknown fragments. Currently, there are 56 SMARTS patterns in the list covering the most used ionizable groups in medicinal chemistry. The algorithm attempts to match every SMARTS pattern on every unknown amino acid. In the case of a match, the corresponding pKa value and the type of equilibrium are added to the list of pKa utilized in eq 1. For each SMARTS pattern, the algorithm keeps track of which atom of the matched substructure is ionizable, i.e., binds or releases a proton. The pKa value is excluded if any previous SMARTS pattern already matched the ionizable atom. Some SMARTS patterns match functional groups with two ionizable centers, for example, a phosphate ion. In such cases, two pKa values and the corresponding types of equilibria are added to eq 1. An initial list of SMARTS patterns and the corresponding pKa values were adopted from Dimorphite-DL,71 where they were derived from experimental data. Later, we adjusted the definitions of most of the SMARTS to improve their use for our particular purpose by removing definitions that were too generic (e.g., aromatic_nitrogen_unprotonated/protonated, primary/secondary/tertiary amines), by adding more functional groups of medicinal chemistry relevance (e.g., tetrazole, imidazole, substituted phenols), or by excluding patterns with pKa outside the relevant pH range (e.g., alcohols with acidic pKa around 14). For the newly added SMARTS patterns, we set the pKa values predicted by ACDlabs. The described SMARTS patterns (see Table S1 of SI) were sufficient to cover the ionizable groups present in 93 non-natural amino acids ready for solid-state peptide synthesis available from the AstraZeneca chemistry inventory.

Constantly Ionized Residues

Some molecular groups have a constantly ionized charge that would not get a predicted pKa value. However, these groups would still affect the net charge of the molecule and the predicted pI as a consequence. In the current implementation, the four-valent nitrogens are identified as constantly ionized.

Sequence-Based pI Calculations

These predictions were performed using a FASTA sequence-based script that is included in the pIChemiSt repository. The sequence-based algorithm is identical to the one of pIChemiSt with the only exception that the canonical amino acids are derived from the single letter FASTA sequences. Both N- and C-termini are assumed to be ionizable. The contribution from the side chains of noncanonical amino acids denoted as X in the sequence is ignored in the calculations. Predictions are reported as mean pI values averaged over the data obtained with the same predefined sets of pKa values of canonical amino acids as in the case of pIChemiSt calculations.

Experimental Data Sets for Validation

Bjerrum et al.46 validated their method on about 100 pI data points of peptides with noncanonical modifications that were retrieved from the Reaxys database.52 We used the Reaxys IDs provided by Bjerrum et al.46 and retrieved the data from Reaxys for our validation. However, while browsing the original references, we discovered that many data points that are attributed as experimental data in the Reaxys database were not measured but calculated. We carefully inspected the sources in Reaxys and discovered that only five publications contained experimental data,72−76 from which we constructed a curated data set. We used the curated data set for validating our software for the case of modified peptides. We also evaluated the tool against the IPC2_peptide assembled set of experimental data of natural peptides as an additional validation.43 The sequences were converted to two-dimensional (2D) structures with ionized N- and C-termini. Structures from the validation set are exemplified in Figure 3.

Figure 3.

Figure 3

Open in a new tab

Representative compounds from the validation set with measured pI data. Blue circles highlight noncanonical modifications: cystine bond,73 β amino acid,72 N-methylation,76 pyroglutamic acid,75 derivative of the tetramethylrhodamine 5-iodoacetamide dye and glycyl-modified lysine side chain.74

Dependencies

pIChemiSt is written in Python3, utilizes RDKit77 for cheminformatics processing and Matplotlib78 for graph plotting. The auxiliary scripts of the distribution utilize BioPython79 functions to read in files in “.fasta” format.

Results

We validated our in-house pKaMatcher against the ACDlabs tool using a set of 271 noncanonical amino acids available for solid-phase synthesis in the AstraZeneca stock. Only 93 amino acids were identified as ionizable by both tools. The correlation between predicted pKa values is shown in Figure 4.

Figure 4.

Figure 4

Open in a new tab

Validation of pKa predictions by pKaMatcher against ACDlabs using a set of 93 ionizable noncanonical amino acids available in the AstraZeneca stock. The coefficient of determination and root-mean-squared deviation are reported at the top of the plot.

We validated pIChemiSt on a curated data set of isoelectric points of modified peptides. The results are shown in Figure 5. In addition, we have validated pIChemiSt on a set of isoelectric points of natural peptides from the IPC2_peptide data set (see Figure 6).43 The sequences from IPC2_peptide set were first converted to chemical structures.

Figure 5.

Figure 5

Open in a new tab

Validation of pIChemiSt on a set of modified peptides. (a) Predictions by the sequence-based model where all noncanonical amino acids and N- and C-termini modifications are ignored; (b) Predictions by pIChemiSt using pKaMatcher as a tool for calculating pKa of noncanonical amino acids; (c) Same as b but with ACDlabs for pKa calculations. The coefficient of determination and root-mean-squared deviation are denoted on the plots. Experimental data are from the following refs (72−76).

Figure 6.

Figure 6

Open in a new tab

Validation of pIChemiSt on IPC2_peptide set of 119093 natural peptides. The coloring is proportional to the density of the points on the plot with lighter color representing denser regions.

Discussion

Importance of Predicting pKa of Noncanonical Amino Acids

The accuracy of pI calculations significantly improves when we introduce automated pKa predictions for noncanonical amino acids and termini modifications. Moving from the plain sequence-based to the structure-based predictions, R2 improved from 0.74 to 0.96 and 0.95 with RMSE shift from 1.09 to 0.44 and 0.51, using ACDlabs and pKaMatcher for pKa predictions, correspondingly. One of the major limitations of sequence-based predictions is that they fail to capture substantial differences between neighboring analogues. For instance, the luteinizing hormone-releasing hormone (LHRH), which has a primary amide at the C-terminus, and its close analogue with a carboxylic acid at the C-terminus, are predicted to have identical pI by the sequence-based approach. However, their pI differs by 2.7 according to the experimental measurements by Heit et al.75 The problem is that the plain FASTA sequence does not allow for decoding the differences in termini of the two peptides. In contrast, the chemical structure-based approach captures the difference and produces an estimation that is close to the experimental value (2.9 as calculated with both ACDlabs and pKaMatcher pKa predictors). Another example regards a case when two dipeptides, α- and β- aspartyl-histidine, appear to have almost identical pI values experimentally. However, their pI values predicted by the sequence-based approach differ by 2.2 units. In particular, the automatically generated FASTA sequences of these peptides, namely, Dh and XH, indicate that the aspartyl side chain is excluded from pI calculations in the case of the β-peptide. In turn, the predictive accuracy greatly improves with the chemical structure-based approach when the difference between pI of α- and β-dipeptides becomes as small as that from the experiments. Such large errors in the predicted pI for very close analogues reflect negatively on isoelectric point calculation methods as differences in pI are largely used to select the pH of mobile phases for efficient chromatographic separation of peptide analogues.24 In fact, if a large pI difference is wrongly predicted, the purification scientist may be misled to believe that the peptides can be separated by varying pH. In turn, when no pI difference is erroneously predicted, the scientist may miss the chance to run a quick and efficient purification. Therefore, peptide purifications are likely to be significantly slowed down by incorrect predictions.

Accuracy of the Predicted pKa Is Less Important than the Misassignment of Ionizable Centers

The chemical structure-based method that uses either ACDlabs or pKaMatcher pKa predictor improves pI predictions substantially as illustrated in Figure 5. The correlation metrics are almost identical for the two pKa predictors, even though the simpler pKaMatcher has much less granularity in capturing substituent effects on the pKa of an ionizable group than the more prominent ACDlabs model. The latter is illustrated by the correlation between the predicted pKa value for a set of ionizable amino acids at AstraZeneca stock that is displayed in Figure 4. The overall correlation statistics is strongly positive, R2 = 0.94 and RMSD = 0.6. However, there is often a spread of points along the Y-axis of ACDlabs pKa compared to constant values along the X-axis of pKaMatcher predictions. The spread is significant and reaches up to 2.5 pKa units. The fact that both methods outperformed the sequence-based approach, where the ionization of noncanonical modifications is ignored, makes us think that the inaccuracy in pKa predictions should be less of a concern for pI calculations than ignoring the ionizable centers completely. However, one should note that the experimental validation set is rather slim and precludes from making solid conclusions.

Model Performs Well for Natural Peptides

We validated the model on a set of natural peptides from the IPC2_peptide43 data set as shown in Figure 6. R2 and RMSE are 0.94 and 0.45, respectively, and they both suggest high accuracy of predictions. The calculations using only the IPC2_peptide pKa set that was trained and validated on this IPC2_data set performed slightly better: R2 and RMSE are 0.97 and 0.25, correspondingly. However, we think that the predictions with different pKa sets are biased to the experimental data from which they were derived. Therefore, we think that it is more instructive to utilize consensus prediction, mean pI values, and the corresponding uncertainty rather than relying on predictions with a specific pKa set.

Limitations of the Described Method

The linear combination of the Henderson–Hasselbalch equations is an approximation. It implies that the ionization state of one group is independent of the ionization states of other groups of the same molecule. It is a valid approximation for remote sites, where electrostatic, van der Waals, and other nonbonded interactions between the groups and electronic induction (I-), mesomeric (M-), and steric hindrance effects can be neglected. For the neighboring ionization centers, the situation becomes more complicated. Theoretically, one should calculate populations of ionization microstates (tautomers of a molecular form with a defined net charge) for each macrostate (a molecular form with a defined net charge). This becomes an exponentially difficult problem for large molecules such as peptides and proteins, and therefore, most of the pKa prediction software limits the size of the molecule they can provide predictions for.51,80 The reader is pointed to a review by Fraczkiewicz on this topic.81 Moreover, pKa of an ionizable group is directly affected by its molecular environment in three-dimensional (3D) space, for example, nonpolar low dielectric regions favor their neutral forms. It implies that conformational flexibility, conformational ensemble, and peptide fold affect pKa of ionizable groups. All the mentioned effects are only indirectly reflected in the current model. The predefined sets of pKa values, trained to reproduce experimental data, average out all of the cases in the training sets, thus, implicitly including the described effects into the pKa values of the side chains and termini. Some pKa sets (e.g., Gauci60 and ProMost48) provide additional granularity introducing dependency of the ionization of N- and C-termini on the side chain. Additionally, there are attempts to capture 3D conformation effect within the proposed pKa set. For example, Toseland et al.57 and Grimsley et al.55 derived pKa from experimental data of folded proteins, while Thurlkill et al.56 measured pKa values of pentapeptides as a model of disordered protein chains. Moreover, there are physics-based models such as PropKa82 that are widely used to assign protonation states of 3D models of proteins and peptides. However, they require 3D conformation as an input and are rather computationally demanding and therefore fall out of the scope of this work. Overall, the approach described in this work, despite its limitations, provides reasonable accuracy for the compounds in the validation set, and thus, it is thought to be useful for peptide chemical optimization and the design of peptide purification experiments.

Summary and Conclusions

In this work, we have described a model for the calculation of the isoelectric point of modified peptides, which we refer to as pIChemiSt. The model accepts a chemical structure as an input and calculates charge versus pH curves, isoelectric point, net charge at pH 7.4, and isoelectric interval, which is the pH interval where the charge of the peptide is negligible. The latter is useful to characterize molecules with poorly defined or uncertain isoelectric points. The model identifies noncanonical amino acids and other fragments, automatically detects ionizable groups, and predicts their pKa. The ACDlabs GALAS model is used by default for the prediction of pKa if the user has access to the software. Alternatively, our pKaMatcher tool can be used for such a purpose, which provides accurate pKa prediction for a set of nCAAs from the AstraZeneca reagent inventory.

We curated a data set of experimental pI data for modified peptides. For many compounds, their calculated pI values were labeled as experimental data in the Reaxys database. We scrutinized the original publication and identified that only 29 data points out of 99 were true measurements. The proposed chemical structure-based isoelectric point model was validated on the curated set of experimental data. We have shown that the chemical structure-based model significantly improves the accuracy of predictions compared to the plain sequence-based model, with R2 being increased from 0.74 to 0.96 and 0.95 and RMSE decreased from 1.09 to 0.44 and 0.51 where the ACDlabs and pKaMatcher tools were used for pKa predictions of noncanonical amino acids, correspondingly. It is worth mentioning that the model correctly predicts differences between close analogues, for example, peptides with a primary amide or free acid at the C-terminus, where the sequence-based model fails.

We think that pIChemiSt can facilitate design and handling of peptide with noncanonical aminoacids, particularly, during the medicinal chemistry optimization and peptide purification.53

Acknowledgments

The authors acknowledge Jonas Boström and Johan Ulander for contributing to pIChemiSt code.

Glossary

Abbreviations

nCAA

noncanonical amino acids

pI

isoelectric point

PPI

protein–protein interactions

PT

peptide therapeutics

IEC

ion-exchange chromatography

FDA

Food and Drug Administration

PTM

post-translational modifications

HPLC

high-performance liquid chromatography

SI

supporting information;

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jcim.2c01261.

  • Defined SMARTS patterns and the raw data of the validation set compounds are listed in the SI (PDF)

Author Contributions

A.I.F. initiated the work, wrote the prototype code, performed the validation; A.I.F. wrote the manuscript with the support of Z.A., G.M.G., S.V.C.; G.M.G. refactored the source code. All authors discussed the results and contributed to the final manuscript.

The authors declare the following competing financial interest(s): A.I.F., S.V.C., Z.A. and G.M.G. are employees of AstraZeneca and own stock options.

Notes

pIChemiSt is available under Apache2 license at https://github.com/AstraZeneca/peptide-tools. We used v1.2.0 of “peptide-tools” distribution, RDKit v2021.03.1, Matplotlib v3.0.3, BioPython v1.79 in this work. Dimorphite_DL v1.2.4 was accessed at https://github.com/UnixJunkie/dimorphite_dl. The code of pIR and pICalculax were accessed at https://github.com/bigbio/pIR and https://github.com/EBjerrum/pICalculax on 10/10/2022. We used version 14.50.122369 of the PhysChem module of Percepta Batch of ACDlabs software released in 2021. The IPC2_peptide data set was downloaded from http://ipc2.mimuw.edu.pl/datasets.html on 21/10/2022.

Supplementary Material

ci2c01261_si_001.pdf (1.3MB, pdf)

References

  1. Muttenthaler M.; King G. F.; Adams D. J.; Alewood P. F. Trends in Peptide Drug Discovery. Nat. Rev. Drug Discovery 2021, 20, 309–325. 10.1038/s41573-020-00135-8. [DOI] [PubMed] [Google Scholar]
  2. Wang L.; Wang N.; Zhang W.; Cheng X.; Yan Z.; Shao G.; Wang X.; Wang R.; Fu C. Therapeutic Peptides: Current Applications and Future Directions. Signal Transduction Targeted Ther. 2022, 7, 48 10.1038/s41392-022-00904-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Passioura T.; Katoh T.; Goto Y.; Suga H. Selection-Based Discovery of Druglike Macrocyclic Peptides. Annu. Rev. Biochem. 2014, 83, 727–752. 10.1146/annurev-biochem-060713-035456. [DOI] [PubMed] [Google Scholar]
  4. Sánchez A.; Vázquez A. Bioactive Peptides: A Review. Food Qual. Saf. 2017, 1, 29–46. 10.1093/fqsafe/fyx006. [DOI] [Google Scholar]
  5. Hamzeh-Mivehroud M.; Alizadeh A. A.; Morris M. B.; Bret Church W.; Dastmalchi S. Phage Display as a Technology Delivering on the Promise of Peptide Drug Discovery. Drug Discovery Today 2013, 18, 1144–1157. 10.1016/j.drudis.2013.09.001. [DOI] [PubMed] [Google Scholar]
  6. Ladner R. C.; Sato A. K.; Gorzelany J.; de Souza M. Phage Display-Derived Peptides as Therapeutic Alternatives to Antibodies. Drug Discovery Today 2004, 9, 525–529. 10.1016/S1359-6446(04)03104-6. [DOI] [PubMed] [Google Scholar]
  7. Adihou H.; Gopalakrishnan R.; Förster T.; Guéret S. M.; Gasper R.; Geschwindner S.; Carrillo García C.; Karatas H.; Pobbati A. V.; Vazquez-Chantada M.; Davey P.; Wassvik C. M.; Pang J. K. S.; Soh B. S.; Hong W.; Chiarparin E.; Schade D.; Plowright A. T.; Valeur E.; Lemurell M.; Grossmann T. N.; Waldmann H. A Protein Tertiary Structure Mimetic Modulator of the Hippo Signalling Pathway. Nat. Commun. 2020, 11, 5425 10.1038/s41467-020-19224-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Wendt M.; Bellavita R.; Gerber A.; Efrém N.-L.; van Ramshorst T.; Pearce N. M.; Davey P. R. J.; Everard I.; Vazquez-Chantada M.; Chiarparin E.; Grieco P.; Hennig S.; Grossmann T. N. Bicyclic β-Sheet Mimetics That Target the Transcriptional Coactivator β-Catenin and Inhibit Wnt Signaling. Angew. Chem., Int. Ed. 2021, 60, 13937–13944. 10.1002/anie.202102082. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Rezhdo A.; Islam M.; Huang M.; Van Deventer J. A. Future Prospects for Noncanonical Amino Acids in Biological Therapeutics. Curr. Opin. Biotechnol. 2019, 60, 168–178. 10.1016/j.copbio.2019.02.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Middendorp S. J.; Wilbs J.; Quarroz C.; Calzavarini S.; Angelillo-Scherrer A.; Heinis C. Peptide Macrocycle Inhibitor of Coagulation Factor XII with Subnanomolar Affinity and High Target Selectivity. J. Med. Chem. 2017, 60, 1151–1158. 10.1021/acs.jmedchem.6b01548. [DOI] [PubMed] [Google Scholar]
  11. Renukuntla J.; Vadlapudi A. D.; Patel A.; Boddu S. H. S.; Mitra A. K. Approaches for Enhancing Oral Bioavailability of Peptides and Proteins. Int. J. Pharm. 2013, 447, 75–93. 10.1016/j.ijpharm.2013.02.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Ding Y.; Ting J. P.; Liu J.; Al-Azzam S.; Pandya P.; Afshar S. Impact of Non-Proteinogenic Amino Acids in the Discovery and Development of Peptide Therapeutics. Amino Acids 2020, 52, 1207–1226. 10.1007/s00726-020-02890-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Lau J.; Bloch P.; Schäffer L.; Pettersson I.; Spetzler J.; Kofoed J.; Madsen K.; Knudsen L. B.; McGuire J.; Steensgaard D. B.; Strauss H. M.; Gram D. X.; Knudsen S. M.; Nielsen F. S.; Thygesen P.; Reedtz-Runge S.; Kruse T. Discovery of the Once-Weekly Glucagon-Like Peptide-1 (GLP-1) Analogue Semaglutide. J. Med. Chem. 2015, 58, 7370–7380. 10.1021/acs.jmedchem.5b00726. [DOI] [PubMed] [Google Scholar]
  14. Kuepper A.; McLoughlin N. M.; Neubacher S.; Yeste-Vázquez A.; Collado Camps E.; Nithin C.; Mukherjee S.; Bethge L.; Bujnicki J. M.; Brock R.; Heinrichs S.; Grossmann T. N. Constrained Peptides Mimic a Viral Suppressor of RNA Silencing. Nucleic Acids Res. 2021, 49, 12622–12633. 10.1093/nar/gkab1149. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Wallraven K.; Holmelin F. L.; Glas A.; Hennig S.; Frolov A. I.; Grossmann T. N. Adapting Free Energy Perturbation Simulations for Large Macrocyclic Ligands: How to Dissect Contributions from Direct Binding and Free Ligand Flexibility. Chem. Sci. 2020, 11, 2269–2276. 10.1039/C9SC04705K. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Valeur E.; Guéret S. M.; Adihou H.; Gopalakrishnan R.; Lemurell M.; Waldmann H.; Grossmann T. N.; Plowright A. T. New Modalities for Challenging Targets in Drug Discovery. Angew. Chem., Int. Ed. 2017, 56, 10294–10323. 10.1002/anie.201611914. [DOI] [PubMed] [Google Scholar]
  17. Gangloff N.; Ulbricht J.; Lorson T.; Schlaad H.; Luxenhofer R. Peptoids and Polypeptoids at the Frontier of Supra- and Macromolecular Engineering. Chem. Rev. 2016, 116, 1753–1802. 10.1021/acs.chemrev.5b00201. [DOI] [PubMed] [Google Scholar]
  18. Lenci E.; Trabocchi A. Peptidomimetic Toolbox for Drug Discovery. Chem. Soc. Rev. 2020, 49, 3262–3277. 10.1039/D0CS00102C. [DOI] [PubMed] [Google Scholar]
  19. Gopalakrishnan R.; Frolov A. I.; Knerr L.; Drury W. J.; Valeur E. Therapeutic Potential of Foldamers: From Chemical Biology Tools To Drug Candidates?. J. Med. Chem. 2016, 59, 9599–9621. 10.1021/acs.jmedchem.6b00376. [DOI] [PubMed] [Google Scholar]
  20. Guéret S. M.; Thavam S.; Carbajo R. J.; Potowski M.; Larsson N.; Dahl G.; Dellsén A.; Grossmann T. N.; Plowright A. T.; Valeur E.; Lemurell M.; Waldmann H. Macrocyclic Modalities Combining Peptide Epitopes and Natural Product Fragments. J. Am. Chem. Soc. 2020, 142, 4904–4915. 10.1021/jacs.0c00269. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Khoury G. A.; Baliban R. C.; Floudas C. A. Proteome-Wide Post-Translational Modification Statistics: Frequency Analysis and Curation of the Swiss-Prot Database. Sci. Rep. 2011, 1, 90 10.1038/srep00090. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Post-Translational Modification. Wikipedia (accessed July 25, 2022), 2022.
  23. D’Hondt M.; Bracke N.; Taevernier L.; Gevaert B.; Verbeke F.; Wynendaele E.; De Spiegeleer B. Related Impurities in Peptide Medicines. J. Pharm. Biomed. Anal. 2014, 101, 2–30. 10.1016/j.jpba.2014.06.012. [DOI] [PubMed] [Google Scholar]
  24. Denton E.How to use the isoelectric point to inform your peptide purification mobile phase pH. https://selekt.biotage.com/peptideblogs/the-isolectric-point-how-to-use-this-to-your-advantage-in-peptide-purification (accessed August 11, 2022).
  25. Proimmune. https://www.proimmune.com/ (accessed August 11, 2022).
  26. Isidro-Llobet A.; Kenworthy M. N.; Mukherjee S.; Kopach M. E.; Wegner K.; Gallou F.; Smith A. G.; Roschangar F. Sustainability Challenges in Peptide Synthesis and Purification: From R&D to Production. J. Org. Chem. 2019, 84, 4615–4628. 10.1021/acs.joc.8b03001. [DOI] [PubMed] [Google Scholar]
  27. Mant C. T.; Hodges R. S. Separation of Peptides by Strong Cation-Exchange High-Performance Liquid Chromatography. J. Chromatogr. 1985, 327, 147–155. 10.1016/s0021-9673(01)81643-5. [DOI] [PubMed] [Google Scholar]
  28. Mant C. T.; Chen Y.; Yan Z.; Popa T. V.; Kovacs J. M.; Mills J. B.; Tripet B. P.; Hodges R. S.. HPLC Analysis and Purification of Peptides. In Peptide Characterization and Application Protocols; Springer, 2007; Vol. 386, pp 3–55. 10.1007/978-1-59745-430-8_1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Sanz-Nebot V.; Benavente F.; Toro I.; Barbosa J. Optimization of HPLC Conditions for the Separation of Complex Crude Mixtures Produced in the Synthesis of Therapeutic Peptide Hormones. Chromatographia 2001, 53, S167–S173. 10.1007/BF02490324. [DOI] [Google Scholar]
  30. Edelmann M. J. Strong Cation Exchange Chromatography in Analysis of Posttranslational Modifications: Innovations and Perspectives. J. Biomed. Biotechnol. 2011, 2011, 936508 10.1155/2011/936508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Bouhallab S.; Henry G.; Boschetti E. Separation of Small Cationic Bioactive Peptides by Strong Ion-Exchange Chromatography. J. Chromatogr. A 1996, 724, 137–145. 10.1016/0021-9673(95)00928-0. [DOI] [Google Scholar]
  32. Shaw K. L.; Grimsley G. R.; Yakovlev G. I.; Makarov A. A.; Pace C. N. The Effect of Net Charge on the Solubility, Activity, and Stability of Ribonuclease Sa. Protein Sci. 2001, 10, 1206–1215. 10.1110/ps.440101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Zapadka K. L.; Becher F. J.; Uddin S.; Varley P. G.; Bishop S.; Gomes dos Santos A. L.; Jackson S. E. A PH-Induced Switch in Human Glucagon-like Peptide-1 Aggregation Kinetics. J. Am. Chem. Soc. 2016, 138, 16259–16265. 10.1021/jacs.6b05025. [DOI] [PubMed] [Google Scholar]
  34. Ouberai M. M.; Dos Santos A. L. G.; Kinna S.; Madalli S.; Hornigold D. C.; Baker D.; Naylor J.; Sheldrake L.; Corkill D. J.; Hood J.; Vicini P.; Uddin S.; Bishop S.; Varley P. G.; Welland M. E. Controlling the Bioactivity of a Peptide Hormone in Vivo by Reversible Self-Assembly. Nat. Commun. 2017, 8, 1026 10.1038/s41467-017-01114-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Zapadka K. L.; Becher F. J.; Gomes dos Santos A. L.; Jackson S. E. Factors Affecting the Physical Stability (Aggregation) of Peptide Therapeutics. Interface Focus 2017, 7, 20170030 10.1098/rsfs.2017.0030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Wang W.; Singh S. K.; Li N.; Toler M. R.; King K. R.; Nema S. Immunogenicity of Protein Aggregates—Concerns and Realities. Int. J. Pharm. 2012, 431, 1–11. 10.1016/j.ijpharm.2012.04.040. [DOI] [PubMed] [Google Scholar]
  37. Åsberg D.; Weinmann A. L.; Leek T.; Lewis R. J.; Klarqvist M.; Leśko M.; Kaczmarski K.; Samuelsson J.; Fornstedt T. The Importance of Ion-Pairing in Peptide Purification by Reversed-Phase Liquid Chromatography. J. Chromatogr. A 2017, 1496, 80–91. 10.1016/j.chroma.2017.03.041. [DOI] [PubMed] [Google Scholar]
  38. Li M.; Josephs R. D.; Daireaux A.; Choteau T.; Westwood S.; Wielgosz R. I.; Li H. Identification and Accurate Quantification of Structurally Related Peptide Impurities in Synthetic Human C-Peptide by Liquid Chromatography–High Resolution Mass Spectrometry. Anal. Bioanal. Chem. 2018, 410, 5059–5070. 10.1007/s00216-018-1155-y. [DOI] [PubMed] [Google Scholar]
  39. Mihailova A.; Lundanes E.; Greibrokk T. Determination and Removal of Impurities in 2-D LC-MS of Peptides. J. Sep. Sci. 2006, 29, 576–581. 10.1002/jssc.200500496. [DOI] [PubMed] [Google Scholar]
  40. Luo H.; Zhong W.; Yang J.; Zhuang P.; Meng F.; Caldwell J.; Mao B.; Welch C. J. 2D-LC as an on-Line Desalting Tool Allowing Peptide Identification Directly from MS Unfriendly HPLC Methods. J. Pharm. Biomed. Anal. 2017, 137, 139–145. 10.1016/j.jpba.2016.11.012. [DOI] [PubMed] [Google Scholar]
  41. Protein Identification and Analysis Tools in the ExPASy Server. https://web.expasy.org/compute_pi/pi_tool-doc.html (accessed July 26, 2022). [DOI] [PubMed]
  42. Kozlowski L. P. IPC – Isoelectric Point Calculator. Biol. Direct 2016, 11, 55 10.1186/s13062-016-0159-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Kozlowski L. P. IPC 2.0: Prediction of Isoelectric Point and PKa Dissociation Constants. Nucleic Acids Res. 2021, 49, W285–W292. 10.1093/nar/gkab295. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. BigBio Stack. 2022, https://Github.Com/Bigbio/PIR (accessed July 26, 2022).
  45. Audain E.; Ramos Y.; Hermjakob H.; Flower D. R.; Perez-Riverol Y. Accurate Estimation of Isoelectric Point of Protein and Peptide Based on Amino Acid Sequences. Bioinformatics 2016, 32, 821–827. 10.1093/bioinformatics/btv674. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Bjerrum E. J.; Jensen J. H.; Tolborg J. L. PICalculax: Improved Prediction of Isoelectric Point for Modified Peptides. J. Chem. Inf. Model. 2017, 57, 1723–1727. 10.1021/acs.jcim.7b00030. [DOI] [PubMed] [Google Scholar]
  47. Skvortsov V. S.; Alekseychuk N. N.; Khudyakov D. V.; Romero Reyes I. V. pIPredict: a computer tool for predicting isoelectric points of peptides and proteins. Biomed. Khim. 2015, 61, 83–91. 10.18097/pbmc20156101083. [DOI] [PubMed] [Google Scholar]
  48. Halligan B. D.; Ruotti V.; Jin W.; Laffoon S.; Twigger S. N.; Dratz E. A. ProMoST (Protein Modification Screening Tool): A Web-Based Tool for Mapping Protein Modifications on Two-Dimensional Gels. Nucleic Acids Res. 2004, 32, W638–W644. 10.1093/nar/gkh356. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Skvortsov V. S.; Alekseychuk N. N.; Miroshnichenko Y. V.; Rybina A. V. PIPredict Version 2: New Features and PTM Analysis. Biomed. Chem. Res. Methods 2018, 1, e00009 10.18097/bmcrm00009. [DOI] [Google Scholar]
  50. Biochemfusion - Downloads. http://www.biochemfusion.com/downloads/#ProteaxDesktop (accessed August 11, 2022).
  51. ACD/Labs . Percepta Platform. https://www.acdlabs.com/products/percepta-platform/ (accessed July 25, 2022).
  52. Reaxys. https://www.reaxys.com/#/search/quick (accessed July 25, 2022).
  53. Peptide-tools code. https://Github.Com/AstraZeneca/Peptide-Tools2022. (accessed July 26, 2022).
  54. Peptide Tools Webserver. http://peptide-tools.com/home (accessed July 26, 2022).
  55. Grimsley G. R.; Scholtz J. M.; Pace C. N. A Summary of the Measured PK Values of the Ionizable Groups in Folded Proteins. Protein Sci. 2009, 18, 247–251. 10.1002/pro.19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Thurlkill R. L.; Grimsley G. R.; Scholtz J. M.; Pace C. N. PK Values of the Ionizable Groups of Proteins. Protein Sci. 2006, 15, 1214–1218. 10.1110/ps.051840806. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Toseland C. P.; McSparron H.; Davies M. N.; Flower D. R. PPD v1.0—an Integrated, Web-Accessible Database of Experimentally Determined Protein PKa Values. Nucleic Acids Res. 2006, 34, D199–D203. 10.1093/nar/gkj035. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Bjellqvist B.; Hughes G. J.; Pasquali C.; Paquet N.; Ravier F.; Sanchez J.-C.; Frutiger S.; Hochstrasser D. The Focusing Positions of Polypeptides in Immobilized PH Gradients Can Be Predicted from Their Amino Acid Sequences. Electrophoresis 1993, 14, 1023–1031. 10.1002/elps.11501401163. [DOI] [PubMed] [Google Scholar]
  59. Bjellqvist B.; Basse B.; Olsen E.; Celis J. E. Reference Points for Comparisons of Two-Dimensional Maps of Proteins from Different Human Cell Types Defined in a PH Scale Where Isoelectric Points Correlate with Polypeptide Compositions. Electrophoresis 1994, 15, 529–539. 10.1002/elps.1150150171. [DOI] [PubMed] [Google Scholar]
  60. Gauci S.; van Breukelen B.; Lemeer S. M.; Krijgsveld J.; Heck A. J. R. A Versatile Peptide PI Calculator for Phosphorylated and N-Terminal Acetylated Peptides Experimentally Tested Using Peptide Isoelectric Focusing. Proteomics 2008, 8, 4898–4906. 10.1002/pmic.200800295. [DOI] [PubMed] [Google Scholar]
  61. Tabb D. L.; McDonald W. H.; Yates J. R. DTASelect and Contrast: Tools for Assembling and Comparing Protein Identifications from Shotgun Proteomics. J. Proteome Res. 2002, 1, 21–26. 10.1021/pr015504q. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Rodwell J. D. Heterogeneity of Component Bands in Isoelectric Focusing Patterns. Anal. Biochem. 1982, 119, 440–449. 10.1016/0003-2697(82)90611-X. [DOI] [PubMed] [Google Scholar]
  63. EMBOSS. http://emboss.sourceforge.net/apps/release/6.6/emboss/apps/iep.html (accessed July 26, 2022).
  64. Nozaki Y.; Tanford C.. [84] Examination of Titration Behavior. In Methods in Enzymology; Enzyme Structure; Academic Press, 1967; Vol. 11, pp 715–734. 10.1016/S0076-6879(67)11088-4. [DOI] [Google Scholar]
  65. Nelson D. L.; Cox M. M.. Lehninger Principles of Biochemistry, 4th ed.; W. H. Freeman: New York, 2004. [Google Scholar]
  66. Solomons T. W. G.Organic Chemistry, Subsequentth ed.; John Wiley & Sons Inc: New York, 1992. [Google Scholar]
  67. http://isoelectric.org/theory.html (accessed July 28, 2022).
  68. http://proteomics.mcw.edu/promost.html (accessed July 28, 2022).
  69. Dwivedi R. C.; Spicer V.; Harder M.; Antonovici M.; Ens W.; Standing K. G.; Wilkins J. A.; Krokhin O. V. Practical Implementation of 2D HPLC Scheme with Accurate Peptide Retention Prediction in Both Dimensions for High-Throughput Bottom-Up Proteomics. Anal. Chem. 2008, 80, 7036–7042. 10.1021/ac800984n. [DOI] [PubMed] [Google Scholar]
  70. Gilar M.; Olivova P.; Daly A. E.; Gebler J. C. Two-Dimensional Separation of Peptides Using RP-RP-HPLC System with Different PH in First and Second Separation Dimensions. J. Sep. Sci. 2005, 28, 1694–1703. 10.1002/jssc.200500116. [DOI] [PubMed] [Google Scholar]
  71. Ropp P. J.; Kaminsky J. C.; Yablonski S.; Durrant J. D. Dimorphite-DL: An Open-Source Program for Enumerating the Ionization States of Drug-like Small Molecules. J. Cheminf. 2019, 11, 14 10.1186/s13321-019-0336-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Greenstein J. P.; Klemperer F. W. Aspartylhistidine. J. Biol. Chem. 1939, 128, 245–250. 10.1016/S0021-9258(18)73747-2. [DOI] [Google Scholar]
  73. Greenstein J. P.; Klemperer F. W.; Wyman J. Further studies on the physical chemistry of cystine peptides. J. Biol. Chem. 1939, 129, 681–692. 10.1016/S0021-9258(18)73632-6. [DOI] [Google Scholar]
  74. Shimura K.; Kamiya K.; Matsumoto H.; Kasai K. Fluorescence-Labeled Peptide PI Markers for Capillary Isoelectric Focusing. Anal. Chem. 2002, 74, 1046–1053. 10.1021/ac0108010. [DOI] [PubMed] [Google Scholar]
  75. Heit M. C.; McFarland A.; Bock R.; Riviere J. E. Isoelectric Focusing and Capillary Zone Electrophoretic Studies Using Luteinizing Hormone Releasing Hormone and Its Analog. J. Pharm. Sci. 1994, 83, 654–656. 10.1002/jps.2600830512. [DOI] [PubMed] [Google Scholar]
  76. Bohner Lang V.; Langguth P.; Ottiger C.; Wunderli-Allenspach H.; Rognan D.; Rothen-Rutishauser B.; Perriard J.-C.; Lang S.; Biber J.; Merkle H. P. Structure–Permeation Raelations of Met-Enkephalin Peptide Analogues on Absorption and Secretion Mechanisms in Caco-2 Monolayers. J. Pharm. Sci. 1997, 86, 846–853. 10.1021/js960387x. [DOI] [PubMed] [Google Scholar]
  77. RDKit: Open-Source Cheminformatics. 2022https://www.rdkit.org (accessed October 10, 2022).
  78. Hunter J. D. Matplotlib: A 2D Graphics Environment. Comput. Sci. Eng. 2007, 9, 90–95. 10.1109/MCSE.2007.55. [DOI] [Google Scholar]
  79. Cock P. J. A.; Antao T.; Chang J. T.; Chapman B. A.; Cox C. J.; Dalke A.; Friedberg I.; Hamelryck T.; Kauff F.; Wilczynski B.; de Hoon M. J. L. Biopython: Freely Available Python Tools for Computational Molecular Biology and Bioinformatics. Bioinformatics 2009, 25, 1422–1423. 10.1093/bioinformatics/btp163. [DOI] [PMC free article] [PubMed] [Google Scholar]
  80. Simulations Plus, Modeling & Simulation Software. https://www.simulations-plus.com/ (accessed September 19, 2022).
  81. Fraczkiewicz R.5.25 - In Silico Prediction of Ionization. In Comprehensive Medicinal Chemistry II; Taylor J. B.; Triggle D. J., Eds.; Elsevier: Oxford, 2007; pp 603–626 10.1016/B0-08-045044-X/00143-7. [DOI] [Google Scholar]
  82. Olsson M. H. M.; Søndergaard C. R.; Rostkowski M.; Jensen J. H. PROPKA3: Consistent Treatment of Internal and Surface Residues in Empirical PKa Predictions. J. Chem. Theory Comput. 2011, 7, 525–537. 10.1021/ct100578z. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

ci2c01261_si_001.pdf (1.3MB, pdf)

Articles from Journal of Chemical Information and Modeling are provided here courtesy of American Chemical Society

RESOURCES