TurboID vs APEX2 vs AP-MS: Choosing the Right Protein Interaction Mapping Strategy

Why Method Choice Is Not Just a Practical Decision

Choosing between TurboID, APEX2, and AP-MS is not primarily a question of lab equipment or protocol familiarity. It is a question of which interactions exist in your data at all.

AP-MS captures proteins that physically copurify with the bait under the lysis and wash conditions used. Interactions that dissociate faster than the experiment takes—transient complexes with half-lives under two minutes, low-affinity assemblies with Kd above 100 µM—are systematically lost, regardless of biological relevance (Huttlin et al., 2021). Proximity labeling takes the opposite approach: the reactive label is applied in living cells, and whatever is within ~10 nm during the pulse is covalently marked. Lysis conditions no longer determine which interactions you find.

The consequence is that AP-MS and proximity labeling recover overlapping but genuinely distinct protein sets. AP-MS identifies proteins in a stable complex; proximity labeling identifies proteins in close spatial proximity to the bait, resulting in distinct protein identifications (Weldatsadik et al., 2021). A 2025 head-to-head study found that TurboID and APEX2 also recover distinct subsets from the same compartment: TurboID enriches membrane proteins and RNA-processing factors, while APEX2 enriches metabolic pathway proteins—differences that arise partly from differential amino acid targeting (Borniger et al., 2025).

For background on TurboID's enzymatic mechanism, see our TurboID proximity labeling principles guide. For comparing TurboID with BioID and miniTurbo, see our proximity labeling enzyme selection guide.

How Each Method Works: Chemistry and Workflow

Mechanistic comparison: AP-MS enriches from lysate; APEX2 and TurboID label proteins in living cells via distinct reactive intermediates.

AP-MS uses an affinity tag on the bait protein to pull down associated proteins from cell lysate. After enrichment, co-purified proteins are trypsin-digested and identified by LC-MS/MS. The entire process occurs post-lysis; only interactions stable enough to survive lysis, dilution, and washing contribute to the final protein list.

APEX2 is an engineered soybean ascorbate peroxidase (~28 kDa). A brief H₂O₂ pulse (30–60 seconds, typically 1 mM) oxidizes biotin-phenol into a phenoxyl radical with a millisecond half-life. This radical reacts with tyrosine and tryptophan residues on proteins within ~10 nm. Labeling is complete in ~1 minute and occurs in intact living cells before any lysis step.

TurboID is an engineered E. coli biotin ligase (~35 kDa). It converts biotin and ATP into reactive biotin-AMP (bioAMP), which labels lysine residues on proteins within ~10 nm. No H₂O₂ or toxic reagent is required—only exogenous biotin (50–500 µM). A 10-minute pulse is sufficient in mammalian cell culture (Branon et al., 2018).

Criterion 1: Interaction Type — Stable vs Transient vs Spatial Neighbor

AP-MS is optimized for stable, high-affinity interactions. Large, well-characterized protein complexes—the ribosome, the nuclear pore, the proteasome—are recovered with high reproducibility by AP-MS. The problem is the biology that happens between stable complexes: kinase-substrate encounters, receptor-adaptor contacts, and chaperone-client associations are transient by design. A kinase that phosphorylates hundreds of substrates cannot maintain stable complexes with all of them simultaneously; the interaction exists for seconds, not minutes.

Tandem affinity purification partially addresses non-specific binding through sequential purification steps, but the stringent washes can inadvertently remove weak interactions—creating an irreducible trade-off between specificity and sensitivity.

TurboID and APEX2 both capture these transient interactions because biotinylation is applied before lysis. Any protein within ~10 nm during the pulse gets labeled, regardless of interaction affinity or lifetime. This is why proximity labeling recovered the postsynaptic density proteome in intact neurons—a structure so extensively crosslinked in situ that standard affinity purification cannot dissociate and re-enrich from it without losing most relevant proteins.

The temporal distinction between APEX2 (~1 min) and TurboID (~10 min) is meaningful when biology moves fast: if you want the proteome within 60 seconds of receptor activation, APEX2 has an advantage. For most cellular processes that unfold over minutes, TurboID’s window is sufficient.

Criterion 2: Toxicity, Cell Viability, and Organism Compatibility

H₂O₂ is APEX2’s fundamental constraint outside cell culture.

The concentrations used for APEX2 labeling (1 mM H₂O₂, 30–60 seconds) cause acute oxidative damage to cellular components. In practice, this restricts APEX2 to mammalian cell culture systems where the H₂O₂ pulse can be terminated rapidly and cells are not needed post-labeling. Biotin-phenol also has limited tissue penetration, making uniform labeling across cell layers or intact tissue technically challenging.

TurboID requires only exogenous biotin—a vitamin with no acute toxicity at working concentrations. This makes TurboID compatible with:

• Primary cell culture, including neurons, cardiomyocytes, and immune cells
• Long-duration or repeated labeling experiments
• Whole organism experiments—Drosophila, C. elegans, zebrafish, mouse, Arabidopsis, rice
• Endogenous-level CRISPR knock-in expressions without overexpression-related toxicity concerns

AP-MS imposes no toxicity constraints on cells during the experiment, but lysis conditions must maintain protein complexes—a constraint that drives a different set of protocol optimization challenges. If your organism grows below 37°C, the choice between APEX2 and TurboID is already made: only TurboID functions at 25°C in Drosophila and C. elegans.

Criterion 3: Membrane Proteins and Insoluble Complexes

Solubilizing integral membrane proteins for AP-MS requires detergents, and detergents disrupt membrane microdomains and peripheral interaction partners. One study showed that a commonly used anionic detergent dissociates up to 75% of interacting partners from a membrane protein subunit during solubilization, underscoring the inherent compromise between effective solubilization and maintaining native interactions.

Proximity labeling bypasses this entirely. Biotinylation occurs in intact cells where membrane topology and lipid environment are preserved. Streptavidin enrichment then captures biotinylated proteins under fully denaturing conditions, recovering membrane-associated proteins with the same efficiency as soluble ones.

The 2025 head-to-head comparison between TurboID and APEX2 found that TurboID specifically recovers more membrane proteins than APEX2 from equivalent compartments, attributing this partly to differential reactivity with membrane-exposed amino acid residues (Borniger et al., 2025). For research centered on receptor complexes, ion channels, transporters, or membrane-scaffolded signaling hubs, TurboID offers the most complete recovery among the three methods.

Criterion 4: Protease Selection and Peptide Recovery Bias

A technical point that rarely appears in method-selection discussions but materially affects data quality when using TurboID.

TurboID biotinylates lysine residues. Trypsin cleaves after lysine and arginine—the standard LC-MS/MS digestion enzyme. A biotinylated lysine is not efficiently cleaved by trypsin, which means peptides containing the biotinylated site are systematically underrepresented in a standard tryptic digest. A 2025 study confirmed this bias and showed it can be partially mitigated by using endoproteinase GluC, though differences persist to some degree (Borniger et al., 2025).

APEX2 biotinylates tyrosines and tryptophans, which are not trypsin cleavage sites, so standard tryptic digests work without this systematic bias. AP-MS is unaffected by biotinylation chemistry entirely.

In practice:

• TurboID + trypsin: accepted for protein-level quantification; underrepresents biotinylated lysine-containing peptides
• TurboID + GluC (or dual-digest): better site-level coverage of biotinylated residues; longer sample prep
• APEX2 + trypsin: no biotinylation-site bias; standard workflow

TurboID labels lysines (trypsin cleavage sites), creating peptide recovery bias correctable by GluC digestion; APEX2 labels tyrosines without tryptic interference.

Decision Framework: Which Method for Which Question?

Method selection decision tree: interaction type, organism system, and pulse duration requirements determine which approach is appropriate.

• Use AP-MS when your protein forms stable, biochemically defined complexes, when you need high-confidence direct interaction data, or when building large-scale networks where reproducibility across labs matters. Our AP-MS service covers quantitative PPI network studies.
• Use TurboID when your bait interacts transiently, when you work in non-mammalian organisms, primary cells, or intact tissue, or when membrane protein neighborhoods are the target. Our TurboID service covers construct design through LC-MS/MS and statistical filtering.
• Use APEX2 when you need a labeling pulse under 2 minutes in mammalian cell culture and H₂O₂ toxicity is manageable, or when recovering metabolic pathway proteins is a priority.

For TurboID experimental design once you have selected the method, see our resource on how to design a TurboID experiment for PPI studies. For data analysis after LC-MS/MS, see our guide on TurboID mass spectrometry data analysis.

Common Misconceptions and Mistakes

“AP-MS and proximity labeling cover the same biology.” They do not. AP-MS recovers stable complexes; proximity labeling recovers the spatial neighborhood. A protein can appear in one dataset and not the other for genuine biological reasons—not because one method failed. Treating negative results in one method as evidence against an interaction is a common interpretive mistake.

“APEX2 is always better than TurboID because it’s faster.” Speed is one dimension of a multi-dimensional choice. APEX2’s H₂O₂ requirement excludes it from primary neurons, whole organisms, and any experiment where a 30-second oxidative pulse cannot be applied uniformly. TurboID’s 10-minute window is sufficient for most biological questions while remaining safe and organism-compatible.

“More candidate proteins means a better proximity labeling experiment.” Candidate count is not a quality metric. A raw enrichment list of 3,000 proteins without controls is less useful than 80 statistically filtered candidates. The value is in the filtered, high-confidence proximal proteome—not the raw list. See our guide on distinguishing true interactors from background for filtering strategies.

“Co-IP is a cheaper alternative to AP-MS.” Co-IP and AP-MS differ primarily in scale and throughput. AP-MS with LC-MS/MS identifies all co-purifying proteins unbiasedly; Co-IP requires knowing what to blot for. For discovery experiments, AP-MS is appropriate. For validating a specific candidate identified by proximity labeling or AP-MS, our Co-IP service is efficient and targeted.

How Creative Proteomics Can Help

Creative Proteomics offers all three approaches as standalone or integrated services. Our TurboID proximity labeling service covers the complete pipeline from construct design through LC-MS/MS and statistical filtering. Our AP-MS service is available for stable complex characterization and large-scale interaction network projects. Our mass spectrometry platform supports both proximity labeling and AP-MS workflows with quantitative proteomics capabilities.

For projects where the right approach is not clear, our scientists can help evaluate your biological question, bait protein, and organism system against each method’s strengths and limitations. Reach out through our online inquiry page.

Frequently Asked Questions

What is the main difference between TurboID and APEX2 proximity labeling?

Both label proteins within ~10 nm in living cells, but they use different chemistry. APEX2 generates phenoxyl radicals from biotin-phenol and H₂O₂, labeling tyrosine and tryptophan residues in ~1 minute. TurboID generates biotin-AMP from biotin and ATP, labeling lysine residues in ~10 minutes without any toxic reagent. APEX2 is faster but restricted to cell culture; TurboID is safer and compatible with whole organisms.

Can TurboID and AP-MS be used on the same bait protein?

Yes, and combining them is often informative. AP-MS identifies stable binding partners that form biochemically defined complexes; TurboID identifies proximity neighbors including transient interactors. The two datasets overlap but are distinct, and integration typically gives a more complete picture of the protein’s molecular context than either method alone.

When should I choose AP-MS over proximity labeling?

AP-MS is the better choice when your protein forms stable, well-defined complexes that survive lysis and affinity purification, when you need high-confidence binary interaction data, or when building large-scale interaction networks where reproducibility across labs is important. AP-MS is also preferred when interactors may be too far away (>10 nm) for proximity labeling to reach.

Does APEX2 work in whole organisms?

APEX2’s requirement for H₂O₂ and biotin-phenol limits in vivo use. Biotin-phenol has limited tissue penetration, and the acute H₂O₂ pulse is difficult to apply uniformly in tissues. TurboID is the more broadly validated and reliable option for in vivo proximity labeling across Drosophila, C. elegans, zebrafish, and mouse.

Why does TurboID recover different proteins than APEX2 from the same compartment?

TurboID labels lysines; APEX2 labels tyrosines and tryptophans. A 2025 study comparing both enzymes in the same HEK293 compartments found that TurboID recovers more membrane proteins and RNA-processing factors, while APEX2 recovers more metabolic pathway proteins (Borniger et al., 2025). The choice of protease for downstream LC-MS/MS adds a further layer of bias.

What is a realistic candidate list size for a well-controlled TurboID experiment?

After statistical filtering (typically fold-change ≥3 over negative control, FDR-adjusted p < 0.05, minimum 3 biological replicates), most TurboID experiments produce 50–300 candidate proximal proteins depending on bait and compartment. Unfiltered raw lists may contain thousands of proteins and are not interpretable without controls.

Can AP-MS detect membrane protein interactions?

AP-MS can, but detergent-based solubilization disrupts membrane microdomains and peripheral interaction partners. One study found that a commonly used anionic detergent dissociates up to 75% of interacting partners from a membrane protein subunit. Proximity labeling, applied in intact cells before lysis, avoids this problem entirely.

Should I use TurboID or Co-IP for validating a specific interaction?

These methods serve different purposes. TurboID is a discovery tool for mapping the proximal proteome; Co-IP is a validation tool for confirming a specific binary interaction. If you have already identified a candidate by TurboID or AP-MS, Co-IP is the appropriate next step.

Related Services

For Research Use Only. Not for use in diagnostic procedures.

Author: CAIMEI LI | Senior Scientist at Creative Proteomics