Biased Agonism
Biased agonism is the ability of different ligands at one receptor to activate its downstream pathways in different proportions, favouring one branch of signalling over another.
A receptor is not a switch with a single output. It couples to several transducers, typically one or more G proteins plus beta-arrestin, and different ligands stabilise slightly different receptor conformations that engage those transducers unequally. Bias is therefore always relative: it is quantified by comparing a test ligand's potency and maximum in two pathways against the same two values for a reference ligand, usually the endogenous one, which yields a ratio of ratios rather than an absolute property.
The idea has been tested clinically more than once. Oliceridine, a mu-opioid agonist selected for strong G protein coupling and weak arrestin recruitment, was approved by the FDA in 2020 for intravenous use in acute pain. An arrestin-biased angiotensin II analogue taken into acute heart failure did not separate from placebo. Carvedilol is often cited as an arrestin-biased beta-blocker discovered after the fact rather than by design. Bias has been described at the GLP-1 receptor too, where reduced internalisation has been proposed as a route to sustained insulin secretion.
Where bias is real it offers something selectivity cannot: one receptor, two consequences, and a chance to keep the effect while dropping the liability. Reported bias also tells you which assay conditions a compound was optimised in, which is worth knowing before comparing it to anything else.
The error that dominates the literature is mistaking system bias for ligand bias. Different pathways have different amplification, so a pathway with large receptor reserve will look more sensitive for every ligand, and a compound that is merely a weak partial agonist can appear biased if only one readout is amplified. Bias measured in one engineered cell line with one readout is a hypothesis, not a mechanism, and it rarely survives translation intact.