English:Aging – Adversarial causal target selection
aiMOOCAging – Adversarial causal target selection
QR-Code
Introduction
This expert colloquium treats aging research as a causal-inference problem rather than a target-discovery contest. Your task is to ask not merely whether a molecular feature changes with age, but whether changing it would alter a clearly defined outcome in a clearly defined population under an intervention that can, at least in principle, be identified.
The course selects two targets from current primary research:
- S-adenosylmethionine availability in aged muscle stem cells: a regeneration target motivated by work showing that intracellular SAM falls in aged murine muscle stem cells, that H3K9 methylation and heterochromatin decline with it, and that restoration of SAM or inhibition of excessive spermidine synthesis can improve regenerative outcomes.
- Microglial cGAS–STING signaling: a brain-aging target motivated by work showing cGAS–STING activation in aged mouse brain, pharmacological and genetic attenuation of age-associated inflammatory and neurodegenerative phenotypes, and microglia-restricted gain-of-function experiments that produce aging-like brain states.
These are research targets, not treatment recommendations. Most of the decisive causal evidence discussed here comes from mice, isolated cells, transplantation, genetic perturbation, or pharmacological experiments. Human translation remains uncertain.
The central rule of the colloquium is: reward robustness rather than novelty. A target earns confidence when different interventions, different outcomes, different experimental systems, and explicit attempts at falsification converge on the same causal structure.
Learning Goals
By the end of this aiMOOC, you should be able to define an intervention-specific estimand, distinguish a total effect from a mediator effect, draw competing causal models, identify negative-control exposures and outcomes, diagnose selection mechanisms, separate direct observations from authors' interpretations, and design a falsification attempt that could genuinely reduce confidence in a favored model.
You should also be able to explain why an age association, a molecular rescue, a transcriptomic shift, and a behavioral improvement are not interchangeable forms of evidence.
The Adversarial Causal-Inference Toolkit
Estimands Before Mechanisms
An estimand is the causal quantity you want to learn. It must specify the intervention, the population, the outcome, and the time horizon.
For a binary intervention A, potential outcomes can be written as Y(1) and Y(0). A simple average treatment effect is:
ATE = E[Y(1) - Y(0)]
In aging biology, the most common error is to speak about the effect of a target without specifying what intervention on that target is meant. Increasing SAM by supplementation, reducing SAM consumption by inhibiting spermidine synthesis, deleting a gene, and changing dietary methionine are not the same intervention. Likewise, inhibiting cGAS, inhibiting STING, blocking TNF downstream, and reducing mitochondrial DNA release are not the same intervention.
An estimand is identifiable from an experiment only under assumptions appropriate to that design. Random treatment assignment can justify exchangeability for the assigned intervention, but does not automatically identify a mediator effect. Genetic models can improve mechanistic specificity while introducing developmental compensation. Pharmacology can improve translational relevance while introducing off-target effects.
Negative Controls
A negative-control outcome should not plausibly be affected through the proposed causal pathway but should share important sources of bias. A negative-control exposure should not causally affect the outcome but should share confounding or measurement structure with the exposure of interest.
A useful negative control can reveal hidden bias, but a null negative control does not prove the absence of bias. In this colloquium, negative controls are treated as diagnostic probes rather than decorative checkboxes.
Selection Mechanisms
Selection can enter before, during, or after an experiment. Examples include survival to old age, eligibility for injury, successful isolation of viable stem cells, gating thresholds, transplantation engraftment, loss of animals before behavioral testing, single-cell quality-control filters, and choosing only analyzable tissue sections.
Conditioning on a variable influenced by both treatment and prognosis can create collider bias. In aging experiments this is especially important because the intervention may itself change survival, cell recovery, or the proportion of cells that pass quality control.
A Three-Layer Reading Rule
For every primary study, keep three statements separate:
Observation: what was directly measured.
Authors' interpretation: the causal or mechanistic conclusion advanced by the paper.
Colloquium hypothesis: the model you currently consider plausible after comparing competing explanations.
The three layers may agree, but they must never be silently merged.
Target One: Restoring SAM Availability in Aged Muscle Stem Cells
Why This Regeneration Target Was Selected
Kang and colleagues reported that aged murine muscle stem cells, or MuSCs, have reduced intracellular SAM, reduced H3K9 di- and tri-methylation, and reduced heterochromatin markers. They manipulated this system in several ways: direct SAM restoration, inhibition of spermidine synthesis, knockdown of spermidine synthase, transplantation of treated or genetically modified aged MuSCs, and double perturbation of spermidine synthase and the H3K9 methyltransferase SUV39H1.
An independent 2024 study reported that reducing MAT2A, the enzyme that synthesizes SAM from methionine, impaired myogenic differentiation and worsened repair after muscle injury, whereas SAM supplementation improved repair-related outcomes. The two studies do not establish a human therapy, but they provide useful orthogonal perturbations around the same metabolic node.
Primary sources:
Kang et al., Nature Metabolism 2024
Xiao et al., Biomolecules 2024
Observations, Interpretation, and Colloquium Hypothesis
| Layer | Statement |
|---|---|
| Observation | In aged murine MuSCs, intracellular SAM and heterochromatin-associated marks were lower than in young cells. Interventions that restored SAM or reduced spermidine synthesis were associated with stronger heterochromatin signals, less damage or cell death in several assays, and better regenerative performance in transplantation or injured-muscle experiments. |
| Authors' interpretation | Age-related excess use of SAM for polyamine synthesis depletes methyl-donor availability, weakens heterochromatin, increases vulnerability to genomic stress, and contributes causally to MuSC aging and regenerative failure. |
| Colloquium hypothesis | A substantial part of the regenerative benefit is mediated by restoration of H3K9-dependent chromatin stability in aged MuSCs, but SAM probably has additional metabolic effects. The target should therefore be treated as a mechanistic node with pleiotropic branches, not as a one-pathway reagent. |
Competing Causal Models for Regeneration
Model R1: heterochromatin-mediation model
Age → increased spermidine-pathway demand → reduced SAM available for methylation → reduced H3K9 methylation and heterochromatin → greater DNA-damage susceptibility and cell loss after activation → impaired MuSC contribution to regeneration.
This model predicts that restoring SAM should improve regeneration, but that the benefit should be reduced or lost if H3K9 heterochromatin cannot be re-established.
Model R2: pleiotropic-metabolism model
Age → altered metabolism.
SAM restoration → multiple methylation, metabolic, redox, translational, and signaling consequences → improved cell survival and regeneration.
Heterochromatin recovery is partly a biomarker or parallel consequence rather than the dominant mediator.
This model predicts that SAM can improve regeneration even when the specific H3K9 heterochromatin route is blocked.
Model R3: survivor-composition model
Treatment changes which MuSCs survive culture, injury, isolation, or transplantation.
The measured post-treatment population therefore contains a higher fraction of intrinsically robust clones. Apparent chromatin rejuvenation and better regeneration are partly caused by selection of cells rather than within-cell reversal of aging.
This model predicts large treatment effects on cell composition, clone survival, or gating probabilities before the final regenerative outcome is measured.
Identifiable Regeneration Estimand
Define the target population as aged mice subjected to a standardized tibialis-anterior injury, or aged MuSCs collected under a pre-specified isolation protocol.
A pragmatic total-effect estimand is:
E[Y(SAM restoration) - Y(vehicle)]
where Y is a pre-specified regenerative outcome such as centrally nucleated myofiber cross-sectional area at seven days after injury, the number of donor-derived regenerated fibers after transplantation, or a functional endpoint at a fixed follow-up time.
For a controlled experiment with assigned treatment, this contrast can be identified as an intention-to-treat effect if treatment assignment is exchangeable and attrition does not become informative. The age effect itself is not identified by this experiment, because young versus old status is not randomized.
A mechanistic estimand asks whether H3K9 methylation mediates the effect. That requires stronger assumptions or an explicit intervention on the mediator. The published Srm plus Suv39h1 double-knockdown experiment is therefore especially informative because it approximates an epistasis test rather than relying only on correlation between SAM and chromatin marks.
Negative Controls for the Regeneration Model
A strong analysis would include several controls:
- Uninjured muscle: regenerative outcomes should not be inferred from normal uninjured fiber size alone; an uninjured tissue outcome can help detect nonspecific hypertrophy.
- Vehicle treatment: controls handling and administration effects.
- Pathway-orthogonal perturbation: direct SAM supplementation and reduction of SAM consumption through spermidine-synthesis inhibition should converge if SAM availability is the operative node.
- Mediator disruption: preventing restoration of H3K9 methylation while restoring SAM is a negative test for the specific heterochromatin-mediation claim.
- Clone-composition control: barcode or lineage-label cells before treatment so that changes in clonal representation can be separated from within-clone improvement.
Selection Mechanisms for the Regeneration Model
Healthy-old-animal selection: animals that survive to an old experimental age are not a random sample of the birth cohort.
Isolation selection: enzymatic dissociation and fluorescence sorting recover some MuSC states more efficiently than others.
Culture selection: ex vivo treatment may preferentially retain cells that tolerate stress.
Transplantation selection: engraftment depends on host niche, cell viability, and cell-cycle state.
Post-injury selection: an intervention that prevents apoptosis changes the composition of the cell pool from which later chromatin and transcriptomic measurements are made.
Analysis selection: excluding damaged fields, low-quality cells, or weakly engrafted samples can induce treatment-dependent selection if exclusion is not blind to outcome.
Regeneration Falsification Attempt
The strongest falsification of Model R1 is an independent, lineage-resolved experiment in aged MuSCs that restores SAM while preventing the proposed H3K9 mediator from recovering.
One design is to isolate aged Pax7-lineage MuSCs, barcode clones before treatment, assign cells to SAM restoration or control, and independently perturb SUV39H1 or another essential H3K9 methylation component. Equal numbers of viable cells would then be transplanted into matched injured recipient muscles. The primary outcome would be donor-derived regenerated fiber contribution at a fixed time. Secondary outcomes would include H3K9me3, DNA damage, apoptosis, clone retention, and cell-cycle state.
Falsification criterion: if SAM restoration still improves regeneration to a similar degree when H3K9 heterochromatin restoration is demonstrably blocked, the strong claim that heterochromatin restoration is the principal mediator is weakened.
Selection safeguard: the analysis should be based on treatment assignment and baseline clone labels rather than only on post-treatment surviving cells.
Target Two: Microglial cGAS–STING in Brain Aging
Why This Brain-Aging Target Was Selected
Gulen and colleagues studied naturally aged mice, aged human cells and tissues, pharmacological STING inhibition, aged Sting1-deficient mice, microglia, hippocampal tissue, and a tamoxifen-inducible microglia-enriched cGAS gain-of-function model. They reported increased cGAMP and STING-pathway activity in aged brain, reduced age-associated inflammatory and degenerative phenotypes after STING inhibition, and aging-like microglial and neurodegenerative phenotypes after microglial cGAS activation.
A 2026 primary study adds an important adversarial alternative upstream model: extracellular vesicles carrying LINE-1 RNA or its reverse-transcribed products were proposed to activate microglial cGAS–STING from a systemic source. That study supports the pathway as a node while challenging any claim that mitochondrial DNA released within old microglia is the unique upstream trigger.
Primary sources:
Observations, Interpretation, and Colloquium Hypothesis
| Layer | Statement |
|---|---|
| Observation | Aged mouse brains showed cGAS–STING-associated activity, including cGAMP and phosphorylated pathway components. STING inhibition in aged mice was associated with reduced inflammatory signatures, less microgliosis, preserved neuronal measures, and improved performance in hippocampal-dependent behavioral tests. Microglia-enriched cGAS gain of function produced inflammatory microglial states, neuronal loss, and impaired memory-related behavior. |
| Authors' interpretation | Mitochondrial DNA released into the cytosol of aged microglia activates cGAS–STING, which drives inflammatory microglial states, neurotoxicity, neuronal loss, and cognitive decline. |
| Colloquium hypothesis | Microglial cGAS–STING is probably a causal amplifier of age-associated neuroinflammation, but the upstream nucleic-acid source may be heterogeneous and partly systemic. The pathway may be causally important even if the mtDNA-only version of the upstream model is false. |
Competing Causal Models for Brain Aging
Model B1: brain-intrinsic mtDNA model
Age → mitochondrial dysfunction in microglia → cytosolic mtDNA → cGAS → cGAMP → STING → inflammatory microglial state and TNF-associated neurotoxicity → neuronal loss and cognitive impairment.
This model predicts that reducing mtDNA escape, cGAS, STING, or a necessary downstream neurotoxic mediator should interrupt the chain.
Model B2: damage-response model
Age-associated neuronal or glial damage → cytosolic DNA and inflammatory signaling.
cGAS–STING activation is mainly a downstream response to pre-existing degeneration. Inhibition reduces inflammatory amplification and behavioral symptoms but is not a primary driver of the aging lesion.
This model predicts incomplete structural rescue when the pathway is blocked after degeneration is established.
Model B3: systemic-signal model
Age-related peripheral tissues → extracellular vesicles or circulating inflammatory cargo → microglial cGAS–STING activation → neuroinflammation.
This model is compatible with cGAS–STING causality but disputes a purely brain-intrinsic upstream origin.
Model B4: behavioral-performance model
STING inhibition improves sickness behavior, locomotion, endurance, motivation, or peripheral inflammatory state. Better performance in learning tasks is therefore partly behavioral rather than a direct restoration of memory circuitry.
This model predicts improvement in cognitive tests without proportional rescue of neuron density, synaptic measures, or cognition-specific assays after controlling for motor and motivational performance.
Identifiable Brain-Aging Estimand
A pragmatic total-effect estimand in naturally aged mice is:
E[Y(H-151) - Y(vehicle)]
where Y is a pre-specified endpoint at a fixed follow-up time, such as CA1 neuronal density, a hippocampal-dependent memory score, or a composite inflammatory state.
This is an estimand for the intervention H-151, not automatically for all possible forms of STING inhibition. If treatment assignment is exchangeable and outcome ascertainment is blinded with negligible differential attrition, the experiment can identify the treatment effect in the studied aged-mouse population.
A stronger mechanistic estimand is the effect of microglia-specific cGAS activity on brain outcomes. Gain-of-function experiments address sufficiency, but the cleanest test of necessity after aging has already developed would be an inducible microglia-specific loss-of-function experiment initiated in old animals.
Negative Controls for the Brain-Aging Model
- Cytosolic genomic DNA: in the mtDNA-specific model, a genomic-DNA species can serve as a source-specific comparison when mtDNA is proposed to rise selectively in the cytosol.
- Peripheral inflammatory markers: in a microglia-restricted activation model, absence of matched peripheral inflammatory activation helps argue against a purely systemic explanation.
- Visible-platform or cue-guided task: a memory experiment should include a task that depends on vision and locomotion but not the same hippocampal memory requirement.
- Open-field locomotion: a behavioral negative-control outcome can test whether apparent memory rescue is largely explained by activity or motivation.
- Type I interferon blockade: if a downstream neurotoxic claim specifically implicates TNF, a branch intervention that blocks type I interferon signaling but does not rescue neuronal survival can help distinguish branches of the pathway.
- Brain-region comparison: a region with low predicted dependence on the proposed circuit can serve as a spatial negative control for a hippocampus-centered mechanism.
Selection Mechanisms for the Brain-Aging Model
Survival to old age: naturally aged mice are selected survivors.
Behavioral-test eligibility: animals with severe motor or sensory impairment may fail to complete cognitive testing, producing informative missingness.
Microglial isolation: dissociation can induce stress and may preferentially recover particular microglial states.
Single-cell filtering: mitochondrial-read thresholds, gene-count thresholds, and integration procedures can alter the representation of fragile or highly activated cells.
Histology field selection: region-of-interest definition can influence apparent neuron or microglia density.
Drug exposure selection: a brain-permeable drug still has systemic effects, so the observed responder population is not restricted to microglia.
Transgenic activation selection: inducible gain-of-function can create stronger or more synchronized signaling than physiological aging, improving mechanistic contrast while reducing naturalistic realism.
The 2026 Systemic Challenge Model
The 2026 Aging Cell study reported that plasma extracellular-vesicle LINE-1 RNA increased with age in a human sample and that old extracellular vesicles could drive microglial cGAS–STING activation and cognitive or inflammatory phenotypes in experimental systems. H151 and inhibition of LINE-1 reverse transcription attenuated reported effects.
For this colloquium, the importance of this paper is not that it proves a unique new aging mechanism. Its importance is adversarial: it supplies a plausible alternative source of cGAS-activating nucleic acid.
If both the 2023 mtDNA model and the 2026 systemic LINE-1 model are partly correct, then cGAS–STING may be a convergence node that integrates several aging-associated danger signals. A robust target can survive disagreement about its upstream trigger.
Brain-Aging Falsification Attempt
The strongest single falsification attempt is a post-onset, inducible, microglia-specific cGAS loss-of-function experiment in naturally aged animals, combined with a downstream rescue.
A rigorous design would induce microglia-specific Cgas deletion only after aging-associated pathway activation is already measurable. Littermate controls would undergo identical tamoxifen exposure. Outcomes would be pre-registered and include CA1 neuronal density, synaptic markers, inflammatory state, hippocampal-dependent memory, locomotion, vision-dependent performance, and peripheral inflammatory markers.
A downstream rescue arm would reactivate STING in the brain or in microglia through a carefully controlled downstream intervention.
Falsification criterion: if verified microglial cGAS suppression does not improve structural or cognitive outcomes despite suppressing the pathway, the claim that microglial cGAS is necessary for ongoing age-associated neurodegeneration is weakened.
Stronger falsification criterion: if downstream STING reactivation fails to restore the phenotype after cGAS loss, the proposed cGAS → STING chain is weakened.
Alternative-model test: if cognitive performance improves but CA1 neuron density, synaptic markers, and cognition-specific readouts do not, the behavioral-performance model gains credibility.
Ranked Evidence Matrix
The following ranking is by causal identification value for this colloquium, not by journal prestige, novelty, effect size, or translational readiness.
| Rank | Experiment or evidence | Target | What it directly identifies or tests | Main strength | Main vulnerability |
|---|---|---|---|---|---|
| 1 | Microglia-enriched inducible cGAS gain of function with STING-pathway rescue in mice | cGAS–STING | Sufficiency of strong microglial cGAS activation for inflammatory states, neuronal loss, and memory-related impairment | Cell-type-focused perturbation with downstream rescue and brain readouts | Gain of function may exceed physiological aging and does not prove necessity in naturally aged microglia |
| 2 | Srm knockdown followed by combined Srm and Suv39h1 perturbation in aged MuSC transplantation | SAM availability and heterochromatin | Whether the regenerative benefit of reducing SAM consumption depends on an H3K9 methylation mechanism | Epistasis-style mediator test is more informative than simple biomarker correlation | Ex vivo manipulation and transplantation create strong selection and niche effects |
| 3 | H-151 treatment in naturally aged mice plus comparison with aged Sting1-deficient mice | cGAS–STING | Necessity-like evidence for STING signaling in age-associated inflammatory, structural, and behavioral phenotypes | Pharmacological and genetic convergence in natural aging | Drug is systemic; germline or long-term knockout can involve developmental or compensatory effects |
| 4 | SAM supplementation or spermidine-synthesis inhibition in old injured muscle | SAM availability | Total effect of the assigned intervention on regenerative outcomes in aged muscle | Direct intervention in an aged regenerative setting with multiple orthogonal manipulations | SAM is pleiotropic; systemic exposure complicates MuSC-specific mediation |
| 5 | MAT2A inhibition with SAM rescue in muscle-repair models | SAM availability | Whether lowering endogenous SAM synthesis impairs repair and whether SAM can counter part of that impairment | Independent laboratory and orthogonal perturbation of the same metabolic node | Includes cell-line and whole-muscle mechanisms that are not identical to aged MuSC heterochromatin |
| 6 | Old extracellular vesicle and LINE-1 experiments with H151 or reverse-transcriptase inhibition | cGAS–STING | Whether a systemic nucleic-acid source can engage the same pathway and generate aging-relevant phenotypes | Strong adversarial challenge to a single-source mtDNA model while retaining the downstream node | Extracellular-vesicle composition, cargo attribution, and model transferability create additional causal ambiguity |
| 7 | cGAS deletion or H151 in 5xFAD Alzheimer-model mice | cGAS–STING | Whether the pathway contributes to disease-model neuroinflammation and cognitive pathology | Independent disease-context convergence | Alzheimer-model evidence is not the same estimand as normal brain aging |
What the Evidence Matrix Does Not Permit
The matrix does not justify claims that either target is clinically effective in humans, that aging has a single molecular cause, or that a molecular signature becoming more youthful necessarily means organismal rejuvenation.
It also does not permit replacing the intervention-specific estimand with a vague claim such as “targeting aging.” A causal conclusion is always attached to a population, intervention, comparator, outcome, time horizon, and set of assumptions.
Strongest Falsification Attempt
Among the two targets, the most demanding falsification design is the post-onset, microglia-specific cGAS necessity experiment because the current evidence already contains pharmacological inhibition, global genetic evidence, natural aging, cell-state profiling, and microglial gain of function.
The decisive protocol would combine:
- Post-onset intervention: induce microglia-specific cGAS loss after an aging phenotype is established.
- Blinded structural outcome: quantify CA1 neuron density and synaptic markers independently of behavioral scoring.
- Behavioral negative controls: include locomotor, visual, motivational, and cue-guided tasks.
- Peripheral negative controls: monitor systemic inflammatory effects to separate central and peripheral pathways.
- Downstream rescue: reactivate STING downstream of cGAS in a controlled brain-localized manner.
- Selection accounting: report every randomized animal, every exclusion, attrition by group, and all single-cell filtering rules.
- Pre-registered stopping rule: define in advance what pattern would count as failure of the cGAS → STING → neurodegeneration model.
A robust model should survive the possibility of being wrong. If an experiment is constructed so that every outcome can be reinterpreted as support, it is not a serious causal test.
Robustness Scorecard Without Hype
Use this qualitative scorecard when reading any aging-target paper:
| Dimension | Stronger evidence | Weaker evidence |
|---|---|---|
| Intervention | Orthogonal genetic and pharmacological perturbations converge | One compound or one correlation |
| Timing | Intervention begins after the phenotype exists | Lifelong perturbation only |
| Cell specificity | Lineage-restricted necessity and sufficiency tests | Whole-body manipulation with inferred cell type |
| Outcomes | Molecular, structural, functional, and behavioral outcomes agree | Only transcriptomic youthfulness |
| Mediation | Mediator is experimentally blocked or rescued | Mediator merely correlates with treatment |
| Negative controls | Controls test alternative explanations | Controls only verify assay operation |
| Selection | Attrition, gating, filtering, and survivor effects are audited | Post-treatment samples are analyzed without selection analysis |
| Replication | Independent laboratories and models converge | Single-laboratory chain of evidence |
| Translation | Human perturbational evidence | Mouse-only association extrapolated to humans |
Primary Research Reading Guide
When reading the SAM studies, pay special attention to the distinction between direct SAM supplementation, Srm perturbation, polyamine inhibition, H3K9 methylation, and regeneration. Each is a different node in the graph.
When reading the cGAS–STING studies, distinguish cGAS from STING, brain from periphery, mtDNA from other nucleic-acid sources, inflammation from neuron loss, and memory from motor performance.
Recommended primary research:
Xiao W. et al. Inhibition of MAT2A Impairs Skeletal Muscle Repair Function. Biomolecules, 2024.
Gulen M.F. et al. cGAS–STING drives ageing-related inflammation and neurodegeneration. Nature, 2023.
Interactive Tasks
Quiz: Test Your Knowledge
Which statement best defines an estimand? (A causal quantity tied to an intervention population outcome and time horizon) (!A list of statistically significant genes) (!A mechanism proposed in a discussion section) (!A molecular marker that changes with age)
What is the selected regeneration target in this colloquium? (SAM availability in aged muscle stem cells) (!Amyloid production in neurons) (!Telomere length in red blood cells) (!Dopamine synthesis in the adrenal gland)
What is the selected brain-aging target? (Microglial cGAS STING signaling) (!Myofiber actin polymerization) (!Pancreatic insulin secretion) (!Renal sodium transport)
Which experiment most directly tests the proposed H3K9 mediation route? (Blocking H3K9 methylation while reducing SAM consumption) (!Comparing old and young animals) (!Measuring body weight) (!Sequencing untreated muscle)
Why is an open-field assay useful in the brain-aging model? (It can help detect locomotor explanations for cognitive test changes) (!It directly measures cGAMP synthesis) (!It quantifies muscle heterochromatin) (!It prevents all forms of selection bias)
What does a microglial cGAS gain-of-function experiment primarily test? (Sufficiency) (!Human clinical efficacy) (!Population prevalence) (!Dietary causation)
Why can post-treatment cell filtering create bias? (Treatment may affect which cells survive or pass filtering) (!Filtering always randomizes the samples) (!Filtering eliminates all confounding) (!Filtering proves mediator specificity)
What does the 2026 LINE-1 extracellular-vesicle study challenge most directly? (The claim that mtDNA is the only upstream source activating cGAS STING) (!The existence of microglia) (!The chemistry of SAM) (!The process of muscle injury)
Which result would most weaken the strong SAM heterochromatin mediation model? (SAM improves regeneration even when H3K9 heterochromatin restoration is blocked) (!SAM raises intracellular SAM concentration) (!Old cells have lower heterochromatin marks) (!Muscle injury activates stem cells)
What is the main purpose of adversarial target selection? (To design evidence that can discriminate competing causal models) (!To maximize novelty regardless of replication) (!To choose the largest molecular fold change) (!To avoid experiments that could refute a hypothesis)
Memory Game
| Estimand | Causal quantity defined by intervention population outcome and time horizon |
| Mediator | Variable on a proposed pathway between intervention and outcome |
| Collider | Variable affected by two or more causes that can induce bias when conditioned on |
| NegativeControl | Probe designed to reveal bias or alternative explanations |
| Sufficiency | Question of whether activating a factor can produce an outcome |
| Necessity | Question of whether removing a factor prevents or reduces an outcome |
Drag and Drop
| Match the correct terms. | Topic |
|---|---|
| SAM restoration | Regeneration intervention |
| SUV39H1 disruption | Heterochromatin mediator challenge |
| Microglial cGAS activation | Sufficiency test |
| Visible platform task | Behavioral negative control |
| LINE-1 extracellular vesicles | Systemic upstream alternative |
...
Crossword Puzzle
| Estimand | What causal quantity specifies the intervention population outcome and horizon? |
| Microglia | Which resident immune cells are central to the selected brain-aging model? |
| Heterochromatin | What compact chromatin state is proposed to recover when SAM availability is restored? |
| Spermidine | Which polyamine is linked to increased SAM consumption in the regeneration model? |
| Collider | What type of variable can induce bias when conditioned on after being caused by two other variables? |
| Falsification | What process deliberately seeks evidence that could overturn a causal model? |
LearningApps
Cloze Text
Open-Ended Tasks
Easy
- Causal vocabulary map: Draw a concept map connecting intervention, estimand, mediator, negative control, selection, necessity, sufficiency, and falsification.
- Target pathway sketch: Create two one-page pathway diagrams, one for SAM and heterochromatin in aged MuSCs and one for cGAS–STING in aged microglia.
- Observation audit: Choose one figure from a primary paper and write three sentences labeled Observation, Authors' interpretation, and Your hypothesis.
- Negative-control brainstorm: Propose two negative-control outcomes for each target and explain what bias each control is intended to detect.
Standard
- DAG comparison: Draw at least three competing directed acyclic graphs for one target and identify which arrows differ between the models.
- Selection audit: Trace every place where animals or cells can be selected out of an experiment from enrollment through final analysis.
- Estimand rewrite: Convert five vague statements such as target X reverses aging into intervention-specific estimands with population outcome and time horizon.
- Replication dossier: Compare the two SAM-related 2024 studies and identify which findings are genuinely independent replications and which involve different biological systems.
Advanced
- Falsification protocol: Write a preregistered protocol for the post-onset microglia-specific cGAS loss-of-function experiment including exclusion rules and negative controls.
- Mediation analysis design: Develop an experimental strategy to separate the total regenerative effect of SAM restoration from the effect mediated specifically through H3K9 heterochromatin.
- Primary literature colloquium: Lead a seminar in which half the group defends the mtDNA-first cGAS model and half defends the systemic LINE-1 model, while both sides must state what evidence would change their view.
- Robustness video: Produce a five-minute scientific video that ranks experimental evidence by causal identification value and explicitly explains why novelty is not the ranking criterion.
Learning Assessment
- Causal model discrimination: Given a new experiment in aged MuSCs, decide which of the competing SAM models it can distinguish and justify your answer in terms of interventions and counterfactual contrasts.
- Negative-control analysis: Evaluate whether a proposed negative control is causally valid or whether the control could itself be affected by the intervention.
- Selection-bias reconstruction: Reconstruct how treatment-dependent cell survival could create an apparent rejuvenation signature even if no individual surviving cell became younger.
- Cross-study triangulation: Explain how pharmacological inhibition, genetic knockout, gain of function, and downstream rescue contribute different information about cGAS–STING causality.
- Translation boundary: Write a short argument specifying what additional human evidence would be required before either target could support a clinical claim.
- Falsification judgment: Given a hypothetical result in which microglial cGAS deletion suppresses inflammatory transcripts but does not rescue neurons or memory, update each competing brain-aging model and explain why.
Evidence of Learning
Evidence of learning should include more than factual recall. A strong learner should be able to produce a causal graph that contains at least one plausible alternative pathway, define an intervention-specific estimand, identify a valid negative-control strategy, explain at least two selection mechanisms, and specify a result that would falsify the favored model.
Practical products can include a preregistration, a causal evidence matrix, an annotated primary-paper figure, a blinded outcome plan, a lineage-tracing or barcoding proposal, a five-minute oral defense, or a replication design.
Transfer is demonstrated when you can apply the same logic to a different aging target, such as senolysis, nutrient sensing, mitochondrial quality control, neurovascular aging, or epigenetic reprogramming, without merely copying the conclusions from this course.
OERs on the Topic
Linked Learning Areas