Human systems / mechanism atlas
The Therapeutic Stack
Small molecule, antibody, and biologic are labels for one layer. A mechanism is the whole stack: what the therapy is, what it recognizes, where it goes, how it enters, what it changes, how it clears, and how the same architecture can fail.
The drug is not the mechanism. It is one component inside a route through a living system.
The vocabulary encourages category errors. Small molecule describes material and scale. Antibody describes a protein architecture. Biologic is an umbrella that includes antibodies. Inhibitor describes an action. None tells you, by itself, whether the therapeutic stays in blood, crosses a membrane, enters a lysosome, changes RNA, recruits immunity, or leaves behind a persistent instruction.
Once the labels are separated into primitives, modalities become comparable. The same target can be occupied, neutralized, destroyed, silenced, or replaced. The same lysosome can be a normal disposal route, an engineered destination, or the site of toxic accumulation. The useful question is never simply “what kind of drug is it?” It is “what sequence of biological events must all be true?”
01 / Format is only the beginning
Six formats, six different stacks
Antibodies are normally catabolized after endocytosis, with FcRn recycling a portion back to circulation; small molecules instead live or die by permeability, metabolism, protein binding, and transport.[1] An antibody-drug conjugate makes the distinction impossible to ignore: antibody, linker, and free payload require separate exposure measurements because each component creates a different mechanism and risk.[2]
Interactive primitive map
Change the format. Watch the whole stack move.
Usually below 1 kDa
Small molecule
A chemically synthesized compound. The format says nothing yet about whether it inhibits, activates, glues, or degrades.
Decisive proof
Free exposure at the site of action, selectivity, metabolite identity, and exposure-response
What makes it selective?
Shape, charge, and noncovalent or covalent binding
Where can it physically go?
Can reach extracellular or intracellular targets if its chemistry permits
How does it cross the cell boundary?
Passive diffusion, transporters, or designed receptor-mediated uptake
What changes when it arrives?
Occupy, inhibit, activate, stabilize, glue, or recruit
How does the system stop?
Metabolism plus renal or biliary elimination
What can the same architecture break?
Off-target binding, reactive metabolites, drug interactions, or lysosomal ion trapping
02 / The compartment that gets blamed
The lysosome is a destination, not a verdict
“It goes to the lysosome” can describe intended clearance or a storage disorder. The route is shared. The failure is not.
Engineered extracellular degraders can bridge a circulating protein to a hepatocyte receptor, inducing endocytosis and lysosomal proteolysis.[3] That is cargo disposal. Phospholipidosis is lipid-processing failure.
Intended lysosomal use
Deliver protein cargo for proteolysis
Receptor capture, endocytosis, acidification, protease action, amino-acid recycling. The molecule may recycle, clear, or be metabolized depending on its design.
Drug-induced phospholipidosis
Trap chemistry and impair lipid turnover
A lipophilic weak base enters, becomes protonated in the acidic lysosome, accumulates, associates with phospholipids, and can inhibit or displace lipid-processing enzymes.[4]
Diffuse
Uncharged, lipophilic weak base crosses a membrane
Protonate
Acidic lysosome converts it to a charged form
Trap
The charged form leaves poorly and exposure rises
Disrupt
Drug-lipid complexes impair phospholipase access or activity
Accumulate
Phospholipids form foamy cells and lamellar bodies
Classic phospholipidosis is therefore most associated with cationic amphiphilic small-molecule chemistry. Large proteins can cause vacuolation or lysosomal accumulation through other routes, but that finding is not automatically phospholipidosis. Even a real phospholipidosis finding must be judged for persistence and adversity: storage without cytotoxicity, inflammation, fibrosis, or organ dysfunction is different from storage with injury.[5]
03 / Toxicity follows architecture
Map the injury to the primitive that created it
“Toxicity” is not a single property carried around by a drug. It is an observed failure that can originate at a different layer from efficacy. The target may be correct while the tissue is wrong; the delivery may be precise while the payload leaks; the vector may reach muscle while immunity makes redosing impossible.
On-target, wrong tissue
The intended biology is harmful when the target also matters in healthy tissue.
Off-target binding
The therapeutic binds something else strongly enough to change physiology.
Distribution injury
The drug reaches an organ that was never supposed to be the main site of action.
Trafficking injury
Material accumulates in endosomes, lysosomes, nuclei, mitochondria, or another vulnerable compartment.
Host-response injury
Innate sensing, complement, cytokines, antibodies, or cellular immunity become part of the pharmacology.
Product-quality injury
Aggregates, impurities, empty particles, free payload, or batch variability change the administered product.
04 / Gene therapy adds a persistent layer
The vector is delivery. The cassette is the instruction.
Gene therapy is easiest to misunderstand when “vector” is treated as the whole product. The actual stack is capsid or carrier, route, cell tropism, intracellular trafficking, genetic cassette, regulatory sequence, transgene, expressed protein, immune response, and durability. Changing any one can change both benefit and risk.
This is why biodistribution is not just pharmacokinetics with a new name. It asks where vector material and expression persist in target and non-target tissues, sometimes for years. FDA guidance also separates biodistribution from shedding, which asks how vector-related material leaves the patient.[6] Genome-editing programs add another proof layer: editing efficiency, unintended edits, chromosomal consequences, and the function of the edited cells.[7]
Capsid / carrier
Which cells can be reached?
Route + dose
Which tissues see how much material?
Promoter / control
Where and how strongly is the cassette read?
Transgene
What protein or editing machinery is produced?
Expression
Is the amount sufficient, excessive, or unstable?
Host response
Can immunity erase benefit or create injury?
Durability
What persists when the original vector particle is gone?
05 / Testing is two axes, not one line
In vitro and in vivo are places. Phases are decisions.
A Phase 3 trial can still contain in vitro assays. A preclinical program can contain sophisticated human tissue. “Later” does not necessarily mean “more biologically realistic” on every axis.
The experimental axis runs from cell-free systems to cells, complex tissue models, animals, and humans. The development axis runs from discovery to preclinical evidence, clinical phases, review, and post-market monitoring.[8] They intersect, but they are not synonyms.
Each rung should retire a particular uncertainty. Phase 1 emphasizes human pharmacology, exposure, and safety; Phase 2 asks for preliminary effectiveness and dose; Phase 3 expands the controlled evidence needed for an overall benefit-risk judgment.[9] Phase 0 microdosing, combined phases, adaptive trials, rare-disease studies, and long-term follow-up change the shape, not the underlying questions.
In silico + chemistry
Could this architecture work, and what liabilities are already visible?
Cell-free / biochemical
Does it bind or catalyze the intended molecular event?
In vitro cells
Can it enter the right cell, reach the right compartment, and change function?
In vivo pharmacology
Does exposure produce the intended biology in a whole organism?
In vivo safety
Which organs fail first, at what exposure, and does the finding reverse?
Phase 0 / exploratory human
Can a very limited human exposure answer an early PK or target-engagement question?
Phase 1 / first-in-human
Can humans tolerate the exposure, and does the mechanism appear?
Phase 2 / proof of concept
Which dose changes disease biology in the intended population?
Phase 3 + confirmation
Does benefit outweigh risk under a prespecified comparative test?
After approval
What appears only with broader use, longer time, or rarer patients?
06 / Competition happens at the seams
A second entrant can keep the job and replace the primitive
Platforms look indivisible until you ask which part actually performs each biological job.
This is also the public-safe way to analyze patents and second entrants: compare architecture, not narratives. Patents may claim compositions, methods, combinations, sequences, linkers, vectors, or manufacturing steps. Know-how may live in optimization and process control. The scientific question is which seam changed and whether the evidence shows an equivalent route through the system.
Recognition
binder, epitope, sequence, or affinity
Routing
receptor, ligand, tropism, cell entry, or compartment
Geometry
linker length, attachment site, valency, or drug-to-antibody ratio
Payload
inhibitor, degrader, toxin, RNA sequence, or transgene
Control
promoter, dose, release logic, reversibility, or recycling
Product
formulation, manufacturing process, purity, stability, and device
The analysis template
Name the cargo. Trace the route. Locate the action. Predict the failure. Match each claim to the experiment that could falsify it.
That sequence works across modalities because it follows the biology rather than the brand name. It also keeps evidence honest: a binding assay proves binding, not tissue delivery; a biomarker proves biological movement, not necessarily clinical benefit; a normal blood test does not exclude a microscopic compartment-level finding.
07 / The system responds
The mechanism does not end when the action succeeds
The stack is not only a route through a living system. It is a wager about what that system will do next.
An intervention can select for escape, resistance, altered routing, or a new distribution of fitness. That response is not outside the mechanism. It changes the next exposure and therefore changes what the same therapeutic architecture means over time.
Bacteria and bacteriophages make the feedback visible because both sides can be sequenced and their interaction can be measured. In one spatially structured E. coli-T7 system, co-evolution produced multiple resistance and infectivity classes, host-range broadening, and repeated genetic changes across the paired populations.[10] But the clean pair is also a warning. A 2025 experiment across three bacteria-phage pairs found that mixed-community conditions could constrain resistance evolution and change whether pairs co-evolved.[11] The map is therefore never just sequence to sequence. It is sequence, partner, environment, time, and measured interaction.
The first prediction asks what the therapy will do. The harder prediction asks what will still be true after the living system answers.
Sources / public evidence only
What this page is built from
This is an educational mechanism map built only from public scientific literature and public FDA guidance. It contains no company-confidential information, no product-specific diligence conclusions, and no medical or legal advice.
- 1Liu (2018), pharmacokinetics of monoclonal antibodies
- 2FDA (2024), clinical pharmacology for antibody-drug conjugates
- 3Caianiello et al. (2021), extracellular protein degradation through hepatocyte uptake
- 4Shayman et al. (2021), lysosomal phospholipase A2 and drug-induced phospholipidosis
- 5Hall et al. (2018), adversity of lysosomal accumulation
- 6FDA (2020), long-term follow-up after gene therapy
- 7FDA (2024), human gene therapy products incorporating genome editing
- 8FDA, the drug development process
- 9FDA, drug development and review definitions
- 10Shaer Tamar & Kishony (2022), multistep diversification in spatial bacteria-phage co-evolution
- 11Castledine et al. (2025), constraints on bacteria-phage co-evolution in a synthetic community