Apoferritin loads a protein cargo by falling apart under acid and reforming once the pH is brought back to neutral — whatever cargo is nearby when the shell closes back up gets swept up inside. Depending on your protein's size, shape, and surface charge, that reassembly plays out one of three ways.

How the call is made: charge is checked first, because a strongly cationic surface can override everything else. Then, for charge-compatible cargos, size decides between Regime I (fits within ~8 nm) and Regime II (larger, flagged for structural flexibility).

1. Is the surface strongly positively charged (net charge above the cationic threshold)? If yes → Regime III.
2. Otherwise, does the protein's longest dimension (Dmax) fit within ~8 nm? If yes → Regime I.
3. If it's charge-compatible but oversized → Regime II, flagged as worth testing since flexibility can't be judged from a static structure alone.

Dmax

The single longest straight-line distance between any two atoms in your protein, in nanometres. This is the strictest test of whether your protein physically fits through the cage's ~8 nm cavity — if Dmax is under 8 nm, your protein's longest "reach" fits inside.

Length × width × thickness

Your protein's bounding box along its three natural axes (longest to shortest), after rotating the structure to align with its own principal axes. This gives a sense of overall shape beyond just Dmax — a rod-shaped and a disc-shaped protein can share a Dmax but have very different proportions.

Volume ratio

The fraction of the cage's internal volume your protein would occupy, treating the cavity as a 268 nm³ sphere (~8 nm across). A ratio of 1.00 would mean your protein alone fills the whole cavity. Even a protein small enough to fit can load less efficiently if it occupies a large share of that space.

Molecular volume

The physical volume of your protein itself, in nm³, before it's expressed as a fraction of the cavity. Estimated with a fast built-in geometric approximation by default, or taken from a ChimeraX solvent-excluded surface if you supplied one under manual overrides.

Net charge (pH 7.4)

Your protein's overall surface charge at physiological pH — estimated with a standard Henderson–Hasselbalch calculation from the sequence, or taken from your own Prot pi value if supplied. This is compared against the cationic threshold (default +5.0) to decide Regime III.

Shape anisotropy (κ²)

A single number from 0 to 1 describing how far your protein departs from a perfect sphere: 0 is perfectly spherical, 1 is a thin rod. Helps you judge whether a large Dmax reflects a genuinely big protein or just an elongated one.

enCAGE reads the heavy atoms (hydrogens excluded) from your structure using Biopython. The coordinates are centred on their centroid and rotated onto their own principal axes — via principal component analysis of the coordinate covariance matrix — which puts the protein's natural long, medium, and short axes into a consistent frame regardless of how the original structure happened to be oriented in the PDB file.

• Length × width × thickness are the extents along these three principal axes.
• Dmax is the largest pairwise distance between any two heavy atoms, computed directly (very large structures are randomly subsampled for speed).
• Shape anisotropy (κ²) comes from the same principal-axis analysis, using the standard relative-shape-anisotropy formula — independent of overall size.

By default, volume is estimated with a fast, dependency-free grid-based approximation: a 3D grid is laid over the structure, and any grid point falling within an atom's van der Waals radius counts as occupied. This is quick to compute but underestimates the true solvent-excluded volume, since it doesn't roll a solvent probe over the surface the way ChimeraX does. If you supply a ChimeraX solvent-excluded surface volume (probe radius 1.4 Å) under manual overrides, that value is used verbatim instead, matching the numbers in the manuscript.

The volume ratio compares this to a fixed 268 nm³ reference cavity (a sphere ~8 nm across) — the estimated internal volume of the equine apoferritin lumen.

By default, net charge is estimated with a standard Henderson–Hasselbalch calculation from the protein's sequence, summing the fractional charge on the N- and C-termini and every ionisable side chain (Lys, Arg, His, Asp, Glu, Cys, Tyr) at the specified pH, using a standard set of pKa values. This is a simplification — it ignores local electrostatic environment and structural burial — so if you have a more rigorous value (e.g. from Prot pi), supplying it under manual overrides will be used instead.

The three descriptors feed into a fixed hierarchy: net charge is evaluated first, because a strongly cationic surface can override everything else and divert the cargo into off-pathway assembly (Regime III). Charge-compatible cargos are then sorted by Dmax against the ~8 nm cavity diameter into Regime I (fits) or Regime II (oversized, flagged for structural flexibility that can't be assessed from a single static structure).

These thresholds — the default pH of 7.4 and the +5.0 cationic cutoff — are calibrated on a small in-house cargo panel (equine apoferritin, pH cycling) described in the accompanying manuscript, and are exposed as adjustable parameters in the tool rather than hard-coded. enCAGE is a heuristic design aid, not a validated classifier: treat Regime II calls especially as "worth testing", not certainty, and expect the quantitative thresholds to need re-tuning for ferritin systems that assemble by a different mechanism — see Examples.

The three proteins used to calibrate enCAGE are built into the homepage as one-click examples — each fetches the structure live from RCSB and runs the full analysis.

Look for "Try an example protein" in the input panel on the homepage.

The same three descriptors were also applied retrospectively to independently published cargo–ferritin systems. Steric compatibility (cavity occupancy) held up cleanly across every system. Electrostatics is more nuanced: it remains the key determinant of outcome everywhere, but the direction and threshold are specific to how a given cage assembles — equine apoferritin reforms passively and is disrupted by very cationic cargo, whereas AfFtn actively co-assembles around cationic cargo, so equine-calibrated regime labels don't transfer directly to it (see footnotes).

^a AfFtn assembles by co-forming around a sufficiently cationic cargo — the opposite mechanism to equine apoferritin's passive pH-cycling. So a cationic cargo flagged as off-pathway (Regime III) by the equine-calibrated rule can encapsulate cleanly in AfFtn, while a cargo that looks charge-compatible (Regime I) by the same rule can fail to encapsulate if it isn't cationic enough to trigger AfFtn's assembly.

^b This acidic cargo is anchored by a cage-specific hydrophobic pocket in Bfr, not by electrostatic complementarity — a reminder that steric fit alone doesn't guarantee the driving mechanism is charge-based.