Guide

Antibody Selection Guide: How to Choose a Research Antibody

A working framework for antibody selection: how to decide between polyclonal and monoclonal, how to think about host species, why epitope choice matters more than most researchers realise, and what validation criteria to check before you spend money on a new reagent.

Browse the antibody catalog →

Antibody selection is the decision framework that determines whether an experiment will produce a clean, interpretable band — or three months of troubleshooting. The wrong antibody can look right in a marketing datasheet and still fail in your system, either because it recognises the wrong isoform, cross-reacts with a paralogue, was raised against a masked epitope, or was validated only in an application unlike yours. Choosing an antibody well means walking through a defined sequence of questions: what target, what application, what species reactivity, what specificity, what validation. This antibody selection guide sets out that sequence.

The framework applies to catalog purchases and to custom antibody projects equally. It also applies retroactively — if a validated antibody is failing in your assay, walking backward through the same questions almost always identifies the mismatch. Most published antibody failures are selection failures, not manufacturing failures. Getting selection right at the start saves reagent cost, animal welfare in downstream in-vivo experiments, and — most importantly — the researcher's time.

Why antibody selection matters

Independent analyses of commercial antibody performance have repeatedly found that a substantial fraction of reagents sold as validated do not detect their nominal target reliably in the application they were bought for. Some studies place the figure above 50%. Even well-cited reagents have been re-evaluated after knockout controls exposed non-specific signal. The consequence is not an abstract reproducibility problem — it is failed grants, unpublishable data, retracted papers, and years of researcher effort applied to the wrong signal.

The root cause is almost always the same: an antibody selection decision was made based on incomplete information — usually a target-and-application match, without checking epitope, without checking paralogue specificity, and without considering whether the validation shown by the manufacturer transfers to the researcher's sample type. A rigorous antibody selection process closes each of these gaps before purchase.

The decision framework — five questions to answer before purchase

Every antibody selection decision reduces to five questions. Answering each explicitly, in order, prevents the most common failure modes.

1. What target — including isoform and functional form

Not just the gene name. Which isoform (there are usually several)? Which processing state (zymogen, active, cleaved, glycosylated)? Which subcellular pool? "MMP-9" is not sufficient — "pro-MMP-9 (92 kDa) secreted form" or "active MMP-9 (82 kDa) after ProMMP-9 activation" is what the antibody must detect.

2. What application

Western blot on reduced denatured samples requires linear epitopes; IHC on FFPE requires epitopes that survive antigen retrieval; IP requires antibody that recognises the native fold; flow cytometry requires membrane-exposed epitopes on intact cells. Antibodies validated in one application do not automatically work in another.

3. What species

Human, mouse, rat, and less common species (zebrafish, Xenopus, primate) have different sequence conservation for different target regions. An antibody raised against a human peptide may recognise the mouse orthologue in a conserved catalytic domain but not in a divergent loop. Check the immunogen sequence against your species of interest.

4. What specificity

Which paralogues does the antibody distinguish? For proteinase families (MMPs, cathepsins, ADAMs, granzymes), the closest paralogue is often more than 60% identical to the target, and pan-family antibodies are common. A specific answer is required before purchase: "recognises MMP-9, does not cross-react with MMP-2, MMP-13, or MMP-1."

5. What validation

What evidence supports the specificity claim? Knockout or knockdown Western blot is the strongest evidence. Recombinant paralogue panel is next. Peptide competition is weaker but still useful. A datasheet showing only a single band on a single lysate is not validation — it is a photograph.

6. Final check — orthogonal application data

Even a well-validated antibody in one application can behave unexpectedly in another. Where possible, prefer antibodies with data from at least two orthogonal applications (e.g., WB + IP, or WB + ELISA). Multi-application validation is a strong indirect signal of reagent quality.

Factors to consider — target, application, host, and clonality

Once the five decision questions are on the table, they resolve into concrete purchasing choices. The main axes of variation are host species, clonality (polyclonal vs monoclonal), immunogen type (peptide vs recombinant protein), and format (unconjugated, conjugated, or purified vs serum).

Host species

Rabbit polyclonal is the default host across most research applications. Rabbit generates high-titre antibodies against most mammalian antigens, rabbit anti-species secondaries are the most commercially available, and rabbit isotype (IgG) is compatible with almost every downstream detection system. Mouse hosts are used almost exclusively for hybridoma-derived monoclonals. Goat and sheep are used when very large quantities of polyclonal are needed (assay development, IP columns). Chicken IgY is used when mammalian Fc cross-reactivity is a problem — IgY does not bind mammalian Fc receptors, does not activate complement, and does not react with rheumatoid factor.

For most researchers selecting a new antibody, the host choice is already made by what is available. When a custom antibody project is being considered, host choice becomes a design decision — see the guides on rabbit polyclonal production and monoclonal vs polyclonal production linked in Pillar Resources below.

Polyclonal vs monoclonal — the fundamental clonality decision

A polyclonal antibody preparation contains a mixture of immunoglobulins from many B-cell clones, each recognising a different epitope on the target. A monoclonal antibody is a single clone recognising a single epitope. This distinction has practical consequences for sensitivity, specificity, and lot consistency.

Polyclonal — broader epitope repertoire

Sensitivity: Higher — multiple epitopes mean higher signal per target molecule. Application robustness: Higher — if one epitope is masked, others compensate. Lot variability: Present — every immunisation campaign produces a slightly different population. Best for: Western blot, IP, most research applications, especially when maximum sensitivity is required.

Monoclonal — defined single epitope

Sensitivity: Lower per binding event but very consistent. Application robustness: Lower — if the single epitope is masked, the antibody fails. Lot variability: None once the hybridoma is established. Best for: Clinical assays, diagnostics, applications requiring defined lot-to-lot consistency, epitope-mapping studies.

Recombinant (any clonality)

Sensitivity: Matches the clonal origin (polyclonal-derived recombinants keep polyclonal breadth; monoclonal-derived stay monoclonal). Lot variability: None — expression from a defined construct. Best for: Long-term studies, multi-lab collaborations, publication reproducibility requirements. See recombinant conversion in Pillar 2.

Superpooled — multi-epitope defined pool

Sensitivity: Very high — combines several domain-specific polyclonals. Application robustness: Very high — multiple epitopes across multiple domains. Best for: Detecting all forms of a proteinase (pro, mature, cleaved), IHC/IF where multi-channel workflow is planned. See the Superpooled Method for the underlying approach.

For proteinase research specifically, the polyclonal-vs-monoclonal decision typically resolves in favour of polyclonal, because most proteinases have multiple functional forms (pro, mature, cleaved) that a multi-epitope reagent detects simultaneously and a single-clone monoclonal cannot. See the proteinase-specific guide linked below.

Epitope considerations — where selection succeeds or fails

Epitope choice is the technical decision that determines whether an antibody will discriminate between paralogues, whether it will detect a specific functional state, and whether it will survive antigen retrieval or reduction. Most antibody failures trace back to an epitope that was not thought about carefully at the selection stage.

Sequence uniqueness — the paralogue problem

For any target with closely related family members, the epitope must fall in a region that is divergent between the target and its paralogues. A peptide antibody raised against a region 90% identical to a paralogue will cross-react. A peptide raised against a region less than 60% identical typically does not. This is why generic "MMP-9 antibody" reagents frequently detect MMP-2 as well — the two enzymes are 65% identical in the catalytic domain, and antibodies raised there will bind both.

Practical rule: before purchase, look up the immunogen sequence on the datasheet. If it is a peptide, BLAST the peptide against your species proteome and specifically against the paralogues you need to distinguish. If it is "full-length recombinant protein," the antibody is likely to have some pan-family reactivity. If the immunogen is not disclosed, treat this as a red flag.

Domain-specific detection

Where a target has multiple functional forms — for proteinases, this is nearly universal — the epitope determines which forms are detected. A propeptide-region antibody detects only the zymogen. A catalytic-domain antibody typically detects both zymogen and mature enzyme (the domain is shared). A C-terminal antibody may distinguish membrane-anchored from shed forms. For any experiment where the functional state matters, the antibody must be selected explicitly for the state of interest.

PTM sites and cleavage sites

Peptides containing phosphorylation, glycosylation, or ubiquitination sites can yield antibodies whose signal is masked when the modification is present. Peptides spanning cleavage sites can yield antibodies that only detect uncleaved forms. Neither is necessarily wrong, but both must be understood in advance — a phospho-masked antibody is exactly the correct choice for detecting the unmodified pool, and a cleavage-site antibody is exactly the correct choice for tracking cleavage. But if your goal is total target detection, PTM- or cleavage-site immunogens are the wrong choice.

Validation criteria — what evidence should you require before purchase

Antibody validation is a hierarchy. Not all validation evidence is equal, and some claims commonly labelled as validation are essentially just photographs. When selecting an antibody, ask which validation methods have been applied and prefer reagents with evidence from higher up the hierarchy.

Strongest evidence — knockout / knockdown

Loss of signal in a genetic knockout, CRISPR-edited cell line, or well-characterised siRNA knockdown is the strongest single piece of evidence that a signal is target-specific. This is now considered the gold standard for validation in many contexts (Uhlen 2016; Roncador 2016). Prefer antibodies with published knockout data whenever possible.

Strong evidence — recombinant paralogue panel

Testing the antibody against a panel of purified recombinant target and closest paralogues — for a proteinase, at least the 2-3 nearest family members — directly demonstrates the specificity claim. This is the standard TPB applies for the catalog.

Moderate evidence — peptide competition

Pre-incubation of the antibody with the immunising peptide abolishes signal. Useful confirmation for peptide-immunogen antibodies but does not rule out cross-reactivity with proteins containing similar epitopes.

Moderate evidence — orthogonal application data

The same antibody detects the same target in two independent applications (e.g., WB and IP; WB and ELISA; IHC and IF). Consistency across methods is a strong indirect signal.

Weak evidence — mass spectrometry immunoprecipitation

The IP eluate is analysed by mass spectrometry and the target is confirmed present. Useful evidence of the target being in the pull-down, but does not confirm exclusive specificity — non-specific interactors are also seen.

Insufficient — datasheet band photograph alone

A single Western blot showing a single band at the expected molecular weight, without any control, is not validation. Almost every antibody produces one dominant band on almost any lysate at some exposure. The correct control is missing.

Post-purchase validation — what to do with the reagent when it arrives

Validation does not end at purchase. Every new antibody must be re-validated in the researcher's own hands, in the exact sample and system where it will be used. Manufacturer validation is a starting point; researcher validation is the actual bar.

  • Titration: Run a dilution series across at least 4-5 dilutions to identify the dilution giving best signal-to-background. Manufacturer-recommended dilutions are frequently sub-optimal. See the dilution optimization protocol.
  • Positive and negative controls: Include a positive control lysate known to contain the target and a negative control lysate lacking it. For proteinases, recombinant purified target is an excellent positive control; a lysate from a knockout cell line is the ideal negative.
  • Loading control: A stably expressed housekeeping protein (GAPDH, actin, tubulin) run on the same membrane confirms equal loading. For proteinases, be aware that some housekeeping controls change abundance under the conditions that also affect proteinase expression.
  • Band size confirmation: Compare the observed band to the expected molecular weight, accounting for glycosylation, propeptide, and post-translational cleavage. For proteinases especially, a band that runs at the "wrong" size may actually be the correct form.
  • Application-appropriate secondary: Use a secondary raised against the primary's host species, at the manufacturer-recommended dilution. Poor secondary choice masquerades as poor primary performance.

Common failure modes and how to avoid them

Certain patterns of antibody failure recur across research groups, and each has a specific root cause at the selection stage.

Failure mode 1 — no signal despite target being present

Most often: the epitope is masked in the researcher's sample. Common causes: reducing conditions cleave a disulphide-dependent epitope; formalin fixation destroyed the epitope in IHC; native fold is required and the sample is denatured. Fix at selection: choose antibodies raised against linear peptides for WB (reduced/denatured), and against native protein or conformational epitopes for IP/native gel.

Failure mode 2 — multiple bands, one at the expected size

Most often: the antibody is genuinely cross-reactive with paralogues, alternative splice forms, or processed products. Fix at selection: check the immunogen sequence and BLAST against the family. If the target has known cleavage products, the multiple bands may all be correct — verify against reference literature.

Failure mode 3 — signal in knockout / knockdown control

Antibody is non-specific. This is the failure mode that gets papers retracted. Fix at selection: prefer antibodies validated with knockout data; if not available, require recombinant paralogue panel data before purchase.

Failure mode 4 — signal works, then stops working in later experiments

Lot-to-lot variability, most commonly with polyclonal antibodies. Fix at selection: for multi-year or multi-lab work, prefer recombinant antibodies (recombinant conversion of polyclonals, or fully recombinant monoclonals). See the custom antibody production pillar for TPB's recombinant conversion option.

Failure mode 5 — antibody works in one lysate, not in another

Sample preparation differences. Frozen vs FFPE, cross-linked vs native, salt concentration in the lysis buffer, protease inhibitors, freeze-thaw cycles, and antigen retrieval conditions all affect epitope accessibility. Fix at selection: match sample preparation to the manufacturer's validation protocol as closely as possible on first use, then optimise from there.

When to consider custom antibody production

Not every experiment can be served by a catalog antibody. Custom production is the appropriate route when the reagent you need does not exist commercially, when existing options have failed in your system, or when long-term supply of a defined-sequence reagent is required.

Signal that custom is the right choice: two or three catalog antibodies for the same target have failed in your system with different modes (no band vs wrong band vs high background); your target has no commercially available antibodies (novel splice variant, disease-associated mutant, less-studied species); you need to distinguish paralogues that no commercial antibody discriminates; you need lot consistency across a multi-year project.

Custom polyclonal antibody projects typically run 18 weeks and cost substantially less than hybridoma-based monoclonal production. Recombinant conversion adds ~6 weeks but produces a defined-sequence reagent equivalent in reproducibility to a recombinant monoclonal at a fraction of the timeline and cost. See the custom antibody production pillar for a full workflow, host-species comparison, and immunogen design considerations.

Antibody selection for proteinase families specifically

Proteinase-family antibody selection has failure modes that don't apply to most other target classes. Every proteinase has a zymogen form and an active form, most have known regulatory cleavage sites, and paralogues within a family typically share more than 60% sequence identity in the catalytic domain. Generic advice ("check the datasheet") is not sufficient — proteinase-specific selection requires attention to functional-state detection and paralogue discrimination that is not needed for, say, a housekeeping protein antibody.

For every major proteinase family, TPB maintains X-vs-Y antibody selection guides that walk through the specific paralogue-discrimination problem for that family (MMP-9 vs MMP-2, cathepsin B vs D, ADAM10 vs ADAM17, and others). These guides are linked in Pillar Resources below and are the appropriate reference for proteinase-antibody selection decisions.

Frequently Asked Questions

What's the single most important factor in antibody selection?

Epitope. Not clonality, not host, not brand — epitope. An antibody raised against the wrong region will fail even if the manufacturer is excellent and the workflow is perfect. Get the immunogen sequence off the datasheet before purchase and think about what forms of your target it will and will not detect.

Should I always prefer monoclonal antibodies for research?

No. Monoclonal antibodies are the correct choice when defined single-epitope specificity and lot consistency are more important than sensitivity or application breadth — clinical assays, diagnostics, epitope-mapping studies. For general research including Western blot, IP, and most IHC, polyclonal or superpooled polyclonal antibodies typically give higher sensitivity and better application robustness. For proteinase research specifically, multi-epitope polyclonal antibodies almost always outperform monoclonals because they detect multiple functional forms simultaneously.

How do I check whether an antibody will distinguish two related paralogues?

Get the immunogen sequence from the datasheet and BLAST it against both paralogues. If the immunogen shares more than 80% identity with the paralogue you want to exclude, cross-reactivity is likely. Less than 60% identity is usually sufficient for discrimination. Also look for datasheet data showing the antibody tested against purified recombinant paralogues side-by-side — this is the strongest evidence of paralogue specificity.

What validation should I do myself before running my main experiment?

Titration in your exact sample type, positive and negative control lysates, a loading control, and — for proteinases especially — a recombinant target run as a size and specificity marker. If knockout or knockdown material is available, run it. Do this validation on a small experimental scale before committing precious samples to the antibody.

How do I know if the manufacturer's validation transfers to my sample type?

Compare the sample type, buffer, and detection system on their validation data to what you'll be using. Human HeLa lysate validation does not automatically transfer to primary mouse tissue lysate. FFPE IHC validation does not automatically transfer to frozen IHC. WB validation says nothing about IP performance. If the validation sample is very different from yours, re-validate in your system before trusting the reagent.

Is knockout validation always necessary?

Knockout or knockdown is the strongest single validation evidence, but it is not always feasible — for essential proteins, for embryonic-lethal knockouts, or for targets in species without established knockout tools. When knockout is not available, the next-best evidence is recombinant paralogue panel testing combined with orthogonal application data. Peptide competition is useful but weaker.

How many controls do I really need in a Western blot?

Minimum: positive control lysate (target present), negative control lysate (target absent or knocked down), loading control (housekeeping protein). For a new antibody, add: a molecular-weight ladder that spans the expected band, an unstained membrane region for background estimation, and if possible a recombinant target run alongside the lysate at a defined concentration. See the controls and validation protocol.

My antibody worked last year and stopped working with a new lot — what do I do?

Titrate the new lot from scratch — polyclonal lots often require a different dilution than previous lots for equivalent signal. If titration doesn't restore signal, compare the CoA between lots (immunogen batch, titre against immunogen, purification method). Contact the manufacturer with lot numbers and your validation data. Longer term, if the target is critical to your work, request a recombinant version of the antibody or start a custom recombinant antibody project — sequence-defined reagents eliminate this failure mode.

Where does citation count fit into antibody selection?

Citation count is useful as a first filter — an antibody with many citations has at least been used successfully in many experiments — but is not sufficient evidence of specificity. Some heavily cited antibodies have been shown to be non-specific once knockout controls were applied. Use citation count as a starting point, then check the specific citations for your application and species, and confirm with your own validation.

When should I stop trying catalog antibodies and start a custom project?

After two or three catalog antibodies for the same target have failed in your system, the failure is probably not random — it is telling you that the commercially available immunogens are wrong for your assay. At that point, a custom project designed with a specific alternative epitope is more likely to succeed than trying more catalog options with the same immunogen strategy. See the custom antibody production pillar for how to design an alternative-epitope project.

Explore related topics in this cluster