This is a great question and one where the field is still actively working out the details. Let me lay out the layers of combinatorial complexity, then map them onto the specific compounds.
The signaling complexity stack
Layer 1: Which Gα subunit gets activated at 5-HT2A itself. The 5-HT2A receptor couples canonically to Gq/11, driving PLC → IP3 + DAG → Ca²⁺ release + PKC. But it’s not just Gq. The 5-HT2AR/mGluR2 heterodimer system shows that hallucinogenic 5-HT2A agonist-induced head-twitch response is regulated by the Gs signaling pathway — meaning some ligands, in the context of the heteromer, can switch G protein coupling entirely, driving cAMP rather than (or in addition to) IP3. There are at least 20 different Gα subunits separated into four main groups, and the specific Gα recruited depends on the conformational state the ligand stabilizes.
Layer 2: Biased agonism — Gq vs. β-arrestin2. This is where the recent literature has converged most clearly. Gq efficacy at the 5-HT2A receptor, not β-arrestin2 recruitment, predicts psychedelic potential, and a threshold level of Gq activation is required to induce psychedelic-like effects. But here’s the subtlety: β-arrestin2 recruitment still matters for the qualitative texture of the experience. LSD stimulates βArr-mediated responses at most tested biogenic amine GPCRs, and LSD-stimulated behaviors in mice require β-arrestin 2 but not β-arrestin 1. So Gq is the gatekeeper for “is this psychedelic at all,” but the ratio of Gq-to-βarr2 signaling shapes what kind of psychedelic experience you get.
Layer 3: The 5-HT2A–mGluR2 heteromer. The Gq/11-coupled 5-HT2A receptor and the Gi/o-coupled mGlu2 receptor assemble into a heteromeric complex in the mammalian brain, and stimulation of these heteromers with an mGlu2 agonist leads to Gq/11 activation by the 5-HT2A partner. This creates bidirectional allosteric crosstalk — the glutamatergic tone in the cortex modulates what the serotonergic ligand does, and vice versa. Three specific residues (Ala-677, Ala-681, Ala-685) at the intracellular end of transmembrane domain four are necessary for mGlu2 to assemble as a heteromer with 5-HT2A, and disruption of this complex attenuates psychosis-like effects of hallucinogenic 5-HT2A agonists. Critically, though, newer research suggests that 5-HT2A and mGlu2 receptors may not physically associate with each other, so this story remains contested — it may be functional crosstalk through shared intracellular signaling rather than literal physical dimerization.
Layer 4: 5-HT2A–D2 heteromers. In the 5-HT2A/D2 heteroreceptor complex, hallucinogenic 5-HT2A agonists like LSD and DOI enhance D2 receptor signaling, while antipsychotics with high 5-HT2A affinity counteract D2 signaling. This is especially relevant for LSD, which has direct D2 affinity on top of its serotonergic action.
Layer 5: Polypharmacology beyond the 5-HT2 family. Each of these compounds hits a different constellation of receptors with different efficacies, and at each receptor the Gα coupling, βarr bias, and heteromer context can differ.
Mapping onto specific compounds
LSD (ergoline): The most polypharmacological of the bunch. High affinity at 5-HT2A, 5-HT2C, 5-HT1A, D1, D2, and adrenergic receptors. Extremely long residence time at 5-HT2A (the “lid” closes over it), which prolongs both Gq and βarr2 signaling. Its D2 activity through the 5-HT2A–D2 heteromer likely contributes to its distinctive stimulant-like edge and thought-acceleration that tryptamines lack. The ergoline scaffold engages amino acid residues in the binding pocket differently from both tryptamines and phenethylamines, producing a unique conformational ensemble.
Psilocin (tryptamine): Structures show psilocin displays a second binding mode at 5-HT2A in addition to the canonical mode, which may underlie its more “organic” feel compared to LSD. More selective for the 5-HT2 family, less D2 involvement, shorter residence time. The dual binding mode may mean it samples a broader conformational landscape at the receptor, producing more variable Gq/βarr2 ratios over time.
DMT (tryptamine): Similar scaffold to psilocin but without the 4-hydroxy group, leading to different binding geometry and notably different kinetics (ultra-rapid onset/offset). DMT has negligible anti-inflammatory efficacy at 5-HT2A unlike many other agonists, suggesting it stabilizes conformations that preferentially activate some downstream pathways while sparing others. Also hits sigma-1 receptors, which none of the others do meaningfully — this may contribute to its distinctive “breakthrough” phenomenology.
5-MeO-DMT (tryptamine): The 5-methoxy substitution dramatically shifts the receptor selectivity profile. Strong 5-HT1A agonism on top of 5-HT2A activity. The 5-HT1A component couples through Gi/o (inhibitory), meaning you get simultaneous Gq activation (via 5-HT2A) and Gi/o activation (via 5-HT1A) in the same cortical neurons. This push-pull between excitatory and inhibitory G protein signaling likely explains the “ego dissolution without visual complexity” phenomenology — massive disruption of self-models without the pattern-generation that Gq-dominant compounds produce. In β-arrestin2 KO mice, the head-twitch response to 5-MeO-DMT is enhanced, suggesting βarr2 normally acts as a brake on 5-MeO-DMT’s effects specifically.
2C-B (phenethylamine): 2C-B shows approximately tenfold greater selectivity for 5-HT2A over 5-HT1A compared to psilocybin, which may bias downstream signaling cascades, and its distinct entactogenic effects may be due to reversal of monoamine transporters. This transporter-reversal activity (releasing serotonin and dopamine) is essentially absent in classical tryptamines and ergolines, adding an MDMA-like floor to the experience. The lipophilicity of 2C-X phenethylamines correlates with their efficacy, with a stronger correlation in the miniGαq assay than the βarr2 assay — meaning the 4-position substituent (bromine in 2C-B) differentially tunes G protein vs. arrestin coupling based on how it fits into a hydrophobic pocket between TM4 and TM5.
2C-I (phenethylamine): Iodine at the 4-position is larger and more polarizable than bromine. 2C-I is listed among the highly potent and fully efficacious anti-inflammatory 5-HT2A agonists, unlike 5-MeO-DMT and DMT which have negligible anti-inflammatory efficacy. This dissociation between anti-inflammatory and psychedelic signaling suggests 2C-I stabilizes a receptor conformation that activates a broader set of downstream effectors than DMT does, even though both are psychedelic. The iodine atom’s size likely forces different TM4-TM5 pocket geometry than bromine, altering the Gq/βarr2 bias ratio.
2C-T-7 (phenethylamine): The propylthio group at position 4 introduces a flexible, sulfur-containing substituent that can adopt multiple conformations within the binding pocket. This conformational flexibility at the receptor level means 2C-T-7 likely samples a wider ensemble of receptor states over time compared to the rigid halogen substituents of 2C-B and 2C-I. Additionally, the thioether metabolizes differently (potential for MAO-mediated toxicity), and the sulfur atom may engage in unique van der Waals contacts with the hydrophobic pocket residues that halogens cannot.
The combinatorial explosion
So for any given psychedelic, the signaling output is determined by:
The specific 5-HT2A conformation stabilized (which depends on scaffold class and 4-position substituent) → which Gα subunit(s) are recruited (Gq, G11, Gs, Gi/o through heteromers) → the Gq/βarr2 bias ratio → the heteromer context (5-HT2A alone vs. 5-HT2A–mGluR2 vs. 5-HT2A–D2) → the parallel activation at other receptor populations (5-HT1A, 5-HT2C, D2, sigma-1, adrenergic) → the kinetics (residence time, onset/offset) → and monoamine transporter effects (2C-x series specifically).
Now, a challenge to the neatness of this framework: most of the biased agonism data comes from heterologous expression systems (HEK293 cells) that don’t recapitulate the heteromer landscape, subcellular compartmentalization, or cell-type-specific Gα subunit expression of actual cortical pyramidal neurons. The bias factors measured in vitro may not map linearly onto in vivo phenomenology. There are three primary scaffolds — tryptamines, ergolines, and phenethylamines — each engaging different subsets of amino acid residues in the receptor binding pocket, and these differences can lead to differential responses for functionally selective outcomes, but we still can’t predict phenomenology from binding data alone. The González-Maeso transcriptome fingerprinting approach (hallucinogenic vs. non-hallucinogenic agonists producing different gene expression patterns) hints that the real discriminator may be at the level of transcriptional programs downstream of these signaling cascades, not the signaling cascades themselves.
Given your molecular docking work with Boltz-2, this might be an interesting case for trying to predict ligand-stabilized receptor conformational ensembles and correlating them with known bias profiles — though the heteromer structures would be the real prize and aren’t yet solved at high resolution.