A recurring line on this site is that an analytical number is only comparable within the method that produced it. HPLC purity depends on wavelength and gradient. EC50 depends on the assay system. Half-life depends on species and route. Those are usually presented as methodological cautions, but underneath a large fraction of them sits one piece of physical chemistry that rarely gets stated directly.
A peptide does not have a charge. A peptide has a charge in a specified solution at a specified pH. Change the solution and you have changed the object — its solubility, its chromatographic retention, its tendency to stick to a container wall, and its propensity to self-associate. The sequence is fixed; the behaviour is not.
This piece is about what you can read off a sequence before anyone runs an instrument, which of those readings are measurements and which are models, and how the answers propagate into documentation you will actually be handed. It is a reading skill, not a handling procedure — no bench instructions appear below.
Where the charges are
A peptide carries ionizable groups at a small, countable set of positions.
- The N-terminus — a free alpha-amino group, protonated and positive at neutral pH.
- The C-terminus — a free carboxyl group, deprotonated and negative at neutral pH.
- Acidic side chains — aspartate (D) and glutamate (E), negative at neutral pH.
- Basic side chains — lysine (K) and arginine (R), positive at neutral pH; arginine's guanidinium group stays protonated across essentially the whole practical pH range.
- Histidine (H) — the interesting one, with a side-chain pKa near 6, meaning it sits partly protonated at physiological pH and flips with small pH shifts.
- Cysteine and tyrosine — ionizable, but only at pH values well above neutral.
Everything else in the twenty-letter alphabet is uncharged. So a first-pass net charge is arithmetic: count K and R as +1, D and E as −1, add +1 for a free N-terminus and −1 for a free C-terminus, treat histidine as a partial and a judgement call.
Run that on the shelf and the results are not uniform:
- BPC-157 (GEPPPGKPADDAGLV) carries one Glu, two Asp and one Lys — net roughly −3 at neutral pH. It is an acidic peptide.
- Epithalon (AEDG) is a four-residue molecule with two acidic side chains — net roughly −2, strongly acidic for its size.
- Selank (TKPRPGP) has a Lys and an Arg and no acidic side chains — net roughly +2.
- LL-37 has eleven basic residues against five acidic ones — net about +6, and that cationic character is not incidental to its story. It is the property that drives association with anionic bacterial membranes.
- MOTS-c is net positive as well, at roughly +3.
Two molecules with the same length, the same purity figure and the same storage condition can therefore be opposite in the one property that governs how they behave in solution.
The isoelectric point is a model, not a measurement
The isoelectric point (pI) is the pH at which the positive and negative charges balance and net charge is zero. It follows directly from which ionizable groups are present and what pKa values you assign them.
That second clause is where the trouble lives. Published pKa sets differ, so calculated pI values differ between tools for the same sequence. Lukasz Kozlowski's Isoelectric Point Calculator work made this explicit: IPC 2.0 (Nucleic Acids Research 2021;49(W1):W285–W292) reports prediction accuracy in RMSD terms — 0.848 versus 0.868 for proteins and 0.222 versus 0.405 for peptides against prior algorithms — which is to say pI prediction is an active modelling problem with quantifiable error, not a lookup.
Three caveats matter more on a research shelf than in a proteomics pipeline:
Terminal modifications change the count. C-terminal amidation removes a negative charge. N-terminal acetylation removes a positive one. A calculator fed bare letters will return the wrong answer for Ipamorelin, Oxytocin, Thymosin Alpha-1 or Gonadorelin — the same failure mode as calculating a theoretical mass from an unannotated sequence, described in the sequence-reading primer.
Non-standard residues are not in the model. Aib, 2-naphthylalanine, dimethyltyrosine and fatty-diacid side chains have no entry in a standard pKa table. A lipidated incretin carries an added carboxyl on its acyl chain that a sequence-only calculator does not see.
A copper complex is a separate axis. GHK-Cu is a tripeptide coordinating a divalent metal ion. Its charge state is not a property of GHK alone.
Amidation is a charge change, not just a stability feature
C-terminal amidation is usually introduced as a protease defence and, in many neuropeptides, a requirement for receptor activity. Both are true. The under-stated third consequence is electrostatic: an amide caps a negative charge that a free acid would carry.
The sharpest illustration on this catalog remains Melanotan 2 versus PT-141, which share an acetylated cyclic core and differ at the C-terminus — amide versus free acid, about one dalton on the mass spectrum. That one-dalton difference is also a one-unit difference in net charge, shifting pI, chromatographic retention and solubility profile alongside whatever it does to receptor engagement. When we called that difference "not a rounding error," this is part of why.
What charge predicts on the paperwork
Net charge is not an abstraction. It sets several things you will see written down.
Counterion load. Basic groups need counter-anions. A peptide with more Lys and Arg binds more trifluoroacetate, which is mass that is not peptide — the mechanism behind low net peptide content figures for basic sequences, covered in the quantitation piece. That is composition, not a quality defect.
Why TFA is there at all. Reversed-phase HPLC of peptides runs in acidic mobile phase for a reason: the acid protonates the peptide and the trifluoroacetate ion-pairs with it, masking charge so the molecule interacts with the C18 phase by hydrophobicity rather than by messy secondary interactions with residual silanols. Åsberg and colleagues (Journal of Chromatography A 2017;1496:80–91) provided direct NMR evidence of peptide–TFA ion-pair formation in acetonitrile–water and showed that elution profiles degrade — a spiked front and long tail — at low ion-pairing-reagent concentration, precisely because the bare peptide ion has strong secondary interactions the ion-paired complex does not.
The consequence is worth stating plainly: a purity percentage is measured on a fully protonated, ion-paired species in an acidic organic mobile phase. That is a legitimate identity and purity condition. It is not the state the molecule occupies in a neutral buffer, and it is another reason the number belongs to its method.
Solubility, and why aggregation clusters near the isoelectric point
Charge is what keeps peptide molecules apart. Like charges repel; the repulsion opposes self-association. Near the pI, net charge approaches zero, repulsion is minimal, and self-association is maximal — which is why solubility for many sequences reaches a minimum in that region and rises as pH moves away in either direction.
Charge is only half of it. The other half is hydrophobicity, conventionally summarised from the Kyte–Doolittle hydropathy scale (Journal of Molecular Biology 1982;157:105–132) as an averaged index across the sequence. Sequences rich in Leu, Val, Ile and Phe self-associate more readily; longer chains have more opportunity to do it; and ordered beta-sheet structure is associated with the worst behaviour, since sheet formation is the structural route to amyloid-type aggregates.
This connects to the degradation-pathways piece, where aggregation appeared as one of four failure modes. It is the only one of the four that is not a covalent chemical reaction — nothing is hydrolysed, oxidised or deamidated. The molecules simply stop being independently dissolved, which is why a purity figure obtained at synthesis says nothing about it.
Prediction here is improving but not solved. The most recent public benchmark is PALM (Eschbach et al., Journal of Chemical Information and Modeling 2026;66(9):5206–5216, from Novo Nordisk's Molecular AI group), a deep-learning model predicting aggregation propensity at single-residue resolution from protein-language-model embeddings. Its own most useful finding is a negative one: PALM failed to identify single mutations that accelerate amyloid-beta aggregation until it was retrained on a much larger dataset. The authors' conclusion is that fine-grained aggregation prediction is still data-limited. Treat any computed aggregation score as a hypothesis about a sequence, not a property of a lot.
The peptide that is no longer in the tube
The most consequential practical result in this area is also the least discussed, and it is a charge phenomenon.
Kristensen, Henriksen and Andresen (PLOS ONE 2015;10(5):e0122419) used analytical HPLC to quantify how much cationic membrane-active peptide adsorbs to the walls of ordinary glass and plastic sample containers. Their conclusion: at typical experimental peptide concentrations, 90% or more of the peptide may be lost from solution through rapid adsorption to container walls.
The mechanism is the same electrostatics discussed above. Borosilicate glass presents anionic silanol groups; a net-positive peptide adsorbs. The peptides tested were mastoparan X, melittin and magainin 2 — not catalog compounds — but the property responsible is net cationic charge, which several compounds on this shelf share.
Three consequences for reading literature:
- Nominal concentration is not achieved concentration in a low-concentration experiment unless someone checked. A reported "1 µM" may be an intended figure.
- It compounds the point made in the evidence-hierarchy piece that concentration is not dose. Here, concentration may not even be concentration.
- It is a plausible contributor to between-lab disagreement on cationic peptides, sitting alongside the sample-preparation artifacts described in the cell-entry explainer — where cell fixation was shown to artifactually redistribute cationic peptides into cells.
None of this is visible on a certificate of analysis. Adsorption is a property of a peptide meeting a surface, not of powder in a vial.
FAQ
Can I get the pI and net charge from the catalog sequence? You can compute an estimate, and the estimate is useful directionally — is this molecule net positive or net negative at neutral pH. But it is only valid if you feed the calculator the fully annotated molecule, including terminal acetylation or amidation, and it will not handle non-proteinogenic residues, acyl chains or coordinated metals at all. See what a sequence line does and does not show.
Does poor solubility mean poor quality? No — and the two questions are answered by different evidence. Solubility behaviour is largely predicted by composition: charge, hydrophobicity, length, structural propensity. A hydrophobic or near-neutral sequence can be entirely correct and still be difficult to work with. Quality questions — identity, purity, counterion, water content, endotoxin — are answered by the documentation described in /quality/. Conflating them leads people to read a physical property as a defect.
Why does this belong in a foundations series rather than a methods post? Because it sits upstream of the methods. Charge state explains why the mobile phase is acidic, why the counterion is trifluoroacetate by default, why basic sequences report lower net peptide content, why aggregation is a stability pathway distinct from the chemical ones, and why low-concentration in vitro numbers can be systematically wrong. One computed property, five downstream consequences — which is a reasonable definition of a foundation. Compound-by-compound sequences and profiles are in the /library/.
This article is educational and for the laboratory research community. Trulogic Labs products are sold for laboratory and research use only and are not for human consumption.