ELISA
ELISA is the workhorse for binding measurements because it is cheap, parallel and needs no specialized instrument. It is also the assay where the format choice quietly determines what the number means, and where most of the trouble comes from the solid phase rather than from the antibody.
The four formats and what each one answers
Direct ELISA coats the antigen and detects with a labeled primary antibody. It has the fewest steps and the fewest places to go wrong, and it requires you to label every antibody you want to test. It suits a small panel you will come back to repeatedly.
Indirect ELISA coats the antigen and detects with an unlabeled primary plus a labeled anti-species secondary. This is the default for screening because nothing has to be labeled. The cost is that the secondary adds its own background and its own species constraint, so a panel mixing human and murine constructs needs two secondaries and cannot be compared across them.
Sandwich ELISA captures the antigen between two antibodies that bind different epitopes. It measures antigen in a complex matrix, which is what you want for quantitation in serum or supernatant. It requires a validated pair, and finding one is a real experiment rather than a formality.
Competition ELISA measures the ability of a soluble competitor to displace binding. This is the format that answers epitope questions and the one that gives you a number comparable to a solution affinity, because the binding event you measure happens in solution rather than on plastic.
Coating is where results are won or lost
Adsorption to polystyrene is passive and denaturing. A protein stuck to a plate is partly unfolded, oriented randomly, and present at a surface density you did not choose. For a conformational epitope, a direct coat can destroy the very thing you are trying to measure, and the antibody that looks weakest may be the one most specific for the native fold.
Capture coating avoids this. Coat with an anti-tag or streptavidin, then capture tagged or biotinylated antigen from solution. The antigen stays folded, orientation is uniform, and surface density is set by the capture step rather than by how sticky the protein happens to be. For any antigen where conformation matters, capture rather than adsorb.
Coating concentration deserves a titration rather than a default. Too little and the signal window closes; too much and you build a dense layer that lets a bivalent IgG bind with both arms, which produces avidity and flatters weak binders.
Blocking, washing and the numbers that follow
Blocking buffer is not interchangeable. Milk contains biotin and immunoglobulins, so it is wrong for any streptavidin or anti-species detection. BSA is the safer default, and for a stubborn background a commercial blocker is usually cheaper than a week of optimization.
Wash stringency sets the apparent affinity. A long, detergent-heavy wash removes fast-dissociating binders and makes the panel look tighter than it is. Keep the wash identical across every plate in a campaign, because a change there moves every number.
Run the antigen-free and primary-free wells on every plate, not once during setup. Background drifts with reagent lots and plate batches, and a control run three weeks ago does not describe today's plate.
What an ELISA number is and is not
An ELISA gives a relative ranking under one set of surface conditions. It supports go or no-go decisions on a panel and it supports quantitation against a standard curve in a sandwich format.
It does not give an affinity. An EC50 from a titration on a coated plate is a function of coating density, valency and wash, and it will not match a solution KD. Use ELISA to reduce a panel from hundreds to tens, then measure the survivors by a solution method before any decision that depends on the absolute number.