From AI derived state signatures to a recovery phase mechanistic explanation
Author: Michael Daniels · Framework: GLA v2.5 · Date: January 19, 2026 · Interpretive synthesis only. Biomarkers associated with PEM are treated as state signatures (predictive), not proof of initiation (causal). This document is not clinical guidance or a treatment recommendation.
Post-exertional malaise (PEM) is the defining symptom of ME/CFS. It reflects a disproportionate, delayed worsening of symptoms after physiological stress, with severity determined primarily by recovery failure rather than exertional intensity itself.
This page uses AI-driven multi-omics findings to examine PEM as a state-linked event—the point at which multiple biological systems fail to return to baseline together. Rather than cataloguing abnormalities at rest, it focuses on biological signals that scale reproducibly with symptom worsening after exertion.
Post-exertional malaise (PEM) is the defining symptom of myalgic encephalomyelitis / chronic fatigue syndrome (ME/CFS), distinguishing it from primary fatigue states and other chronic illnesses. PEM reflects a disproportionate, delayed worsening of symptoms following physiological stress, with severity determined more by recovery failure than by exertional intensity itself.
Because PEM captures the point at which multiple biological systems fail to return to baseline, it provides a uniquely informative window into disease mechanism. Rather than mapping isolated abnormalities at rest, this page focuses on PEM as a state-linked event, using AI-driven multi-omics modeling to identify reproducible biological patterns that scale with symptom worsening after exertion.
This document therefore maps how PEM is generated, not by proposing a single causal driver, but by integrating observed state signatures into a coherent, recovery-phase mechanistic chain. The goal is to explain why exertion appears tolerated initially, why symptoms are delayed, and why recovery becomes progressively more difficult over time.
Post-exertional malaise (PEM) is associated with a coordinated, multi-domain biological state, rather than a single driver or isolated pathway.
Using AI-driven multi-omics modeling, PEM severity was linked to state-dependent changes across multiple biological domains, which rise and fall together and correlate with symptom worsening after exertion.
↓
The immune and molecular signatures identified in association with PEM define a reproducible biological state that co-occurs with symptom severity, but they do not, on their own, establish where PEM is generated or how it unfolds temporally.
These signals describe what is present during PEM, not what initiates it. They therefore require mechanistic interpretation rather than direct causal assignment.
When recovery and clearance are impaired, endothelial and barrier interfaces remain partially unresolved during the recovery phase. These unresolved targets continue to present stress-associated cues, maintaining immune visibility beyond the intended termination window.
As a result, immune programs do not fully disengage. Instead, they remain in a monitoring and containment state, consistent with prolonged interface surveillance rather than active attack.
The immune features observed during PEM—particularly IFN-γ–leaning cytotoxic programs associated with MAIT and γδ T cells—align with this interpretation. These cell types are specialized for sensing stressed or metabolically abnormal interfaces and are well suited to sustained surveillance roles.
Importantly, there is no evidence of:
Immune signals observed during PEM report a recovery-phase control failure elsewhere in the system.
They function as state markers of unresolved stress and incomplete termination, linking observable biomarkers to downstream symptom amplification without reversing causality or assigning immune primacy.
The mechanistic chain below is therefore ordered by physiological timing and recovery logic, tracing how exertional stress is converted into delayed persistence rather than assigning causality to downstream state markers.
Execution → Recovery → Persistence (PEM)
GLA v2.5Figure 1. PEM at a glance — execution, recovery, persistence. This diagram illustrates timing and coordination, not causality. Exertion functions as a stress test that exposes a vulnerability; the biological failure that defines PEM unfolds during recovery, not during exertion itself.
Execution -> Recovery -> Persistence
Physiological shear stress generated during physical or cognitive exertion unmasks a pre-existing control vulnerability, rather than producing immediate tissue failure or energy collapse.
Skeletal muscle functions as the dominant execution surface because it undergoes the largest dynamic perfusion shifts in the body and therefore places the greatest demand on precise microvascular flow regulation. During exertion, global oxygen delivery and cardiac output are preserved, but local flow distribution within muscle becomes increasingly heterogeneous. Capillary recruitment and oxygen extraction lose precision, revealing impaired execution-surface control rather than insufficient supply.
This mismatch does not immediately force failure. Instead, it encodes stress into the system while performance may still appear intact. Exertion therefore functions as a stress test, not a trigger: it exposes a latent instability that becomes biologically consequential only during the recovery phase that follows.
Figure 2 - Execution surface focus: skeletal muscle under shear
Section 1Figure 2. Execution surface focus - skeletal muscle under exertional shear. This diagram illustrates timing and coordination, not causality. It explains why muscle is the dominant execution surface for PEM framing without invoking weakness or primary fatigue: the largest perfusion swing makes distribution precision the limiting factor.
Figure 2B - Execution surface focus: capillary recruitment grid under exertional shear
Section 1Figure 2B. Execution surface focus - capillary recruitment under exertional shear. This diagram illustrates timing and coordination, not causality. It shows how skeletal muscle can experience a distribution precision failure under load even when global delivery is preserved, making exertion a stress test that exposes an execution-surface control vulnerability.
Following exertion, effective recovery requires rapid reset of membrane architecture and clean termination of signaling across metabolic, endothelial, and neuromuscular interfaces.
PI/PLC-linked signaling observed during post-exertional malaise reflects heightened membrane resolution and recycling demand, not pathological receptor hyperactivation. These pathways scale with the intensity and duration of recovery burden, indicating that the system is actively attempting to disengage execution surfaces and restore baseline signaling geometry.
When this coordination is incomplete, signals decay more slowly than intended. Receptor platforms persist beyond their normal window, and recovery processes overlap instead of resolving sequentially. This phase marks the critical transition from exertion to delayed pathology, establishing the conditions under which persistence emerges if termination mechanisms fail.
During recovery, control failure emerges at the level of ER–mitochondrial coordination, rather than from excessive metabolic load or energy demand.
Calcium signaling between the endoplasmic reticulum and mitochondria becomes mis-timed, particularly at mitochondria-associated membranes (MAMs). Instead of brief, well-gated calcium pulses that support signal termination, repair initiation, and recovery sequencing, calcium transfer becomes prolonged, repetitive, or poorly synchronized.
This represents a failure of termination, not calcium overload. The system remains biochemically locked in a partial recovery state, continuing to signal “repair in progress” without achieving closure. As a result, downstream recovery programs remain active beyond their intended window.
This mis-timed Ca²+ routing is the first point at which recovery-phase stress is converted into persistent intracellular signaling, allowing recovery demands to propagate forward even after exertional demand has fully ended.
Mitochondria initially respond appropriately to recovery demands but do so under progressively constrained conditions.
Mis-timed calcium signaling during recovery drives reactive oxygen species generation, which preferentially impacts cardiolipin, a structural lipid essential for inner mitochondrial membrane organization and efficient electron transport. To preserve function, cardiolipin must be repeatedly remodeled during recovery.
When recovery is incomplete, this remodeling becomes recurrent and energetically expensive. Electron transport capacity is maintained, but termination efficiency and structural resilience decline with each cycle.
In this framework, NAD+ dysfunction reflects lipid-coupled misutilization rather than early depletion: NAD+ may remain available, but unstable membranes and impaired ER–mitochondrial (MAM) coupling prevent its productive discharge into recovery processes.
Over successive exertional–recovery events, mitochondrial function is not abruptly lost. Instead, recovery bandwidth gradually erodes. Energy production remains possible, yet it becomes less flexible, less resilient, and more susceptible to disruption by subsequent stress.
This progressive constraint explains why PEM severity and duration increase over time, even when individual exertional exposures are similar and baseline metabolic measures remain relatively preserved.
Execution → Recovery → Persistence
As mitochondrial recovery bandwidth becomes constrained, sphingolipid resolution stalls during the recovery phase, particularly at membrane domains responsible for signal termination and interface reset.
Instead of being fully dismantled and recycled, stress-conditioned sphingolipid structures persist in intermediate states. Ceramide- and HexCer-enriched membrane configurations remain partially intact, sustaining permissive signaling surfaces that favor prolonged receptor residency and delayed disengagement.
This is not continued stimulation. It is failed resolution: membrane architecture remains biased toward persistence rather than closure, even after the initiating stressor has passed.
As a result, the system remains biochemically set to “recovery in progress,” despite normalization of energetic demand and absence of ongoing exertion.
Normal recovery requires a final biochemical completion signal that marks cellular interfaces as resolved and safe to disengage.
In post-exertional malaise, phosphatidylserine (PS) asymmetry is restored inconsistently. Instead of being reliably re-internalized, PS remains intermittently exposed at affected membranes, indicating that recovery has not fully concluded even in the absence of ongoing damage.
As a result, these interfaces continue to present as partially unresolved, preventing full disengagement of downstream recovery and monitoring programs. The system behaves as though repair is incomplete, despite normalization of upstream demand.
This intermittent failure of completion creates a biological memory of stress, lowering the threshold for persistence on subsequent exertion and allowing each recovery episode to become longer and more difficult to resolve.
When completion cues remain unreliable, immune sentinel systems remain engaged during recovery. Mistimed calcium signaling and residual reactive oxygen species at endothelial and barrier surfaces maintain interface visibility, sustaining immune monitoring signals without provoking inflammation.
This engagement is dominated by MAIT and γδ T cell–associated programs, which are specialized for surveillance of stressed interfaces rather than initiation of broad inflammatory responses. Their activity reflects continued visibility of unresolved endothelial and barrier surfaces.
These immune signals do not indicate autoimmune attack or cytokine excess. They represent ongoing surveillance and containment, scaled to the duration and severity of incomplete recovery. As long as interfaces remain flagged as unresolved, immune programs fail to disengage fully.
A prolonged recovery state carries a measurable physiological cost.
Nitrogen-related signals rise in proportion to recovery duration and inefficiency, reflecting sustained protein turnover, redox buffering, and ongoing repair demand rather than primary metabolic or urea-cycle failure. These signals track how long recovery remains incomplete, not whether exertion occurred.
Genetic analyses indicate that multiple variants associated with nitrogen handling, clearance efficiency, and recovery bandwidth may predispose individuals to prolonged repair costs, shaping susceptibility to ME/CFS and influencing PEM duration rather than initiating the disease process.
These downstream costs shape the depth and duration of PEM, but they do not initiate it. They accumulate as a consequence of persistent recovery demand, explaining why PEM becomes longer and more severe with repeated exertion even when activity levels remain similar.
This final layer represents the visible metabolic expense of an unresolved recovery process, not its underlying cause.
In summary, post-exertional malaise arises when physiological shear stress exposes a fragile execution surface in skeletal muscle, initiating a recovery process that fails to terminate cleanly. Exertion increases membrane coordination demand, engaging PI/PLC-linked remodeling and ER–mitochondrial calcium exchange at mitochondria-associated membranes. When calcium timing fails to resolve, recovery-phase reactive oxygen species accumulate, impairing cardiolipin remodeling and progressively reducing mitochondrial recovery bandwidth. As recovery efficiency erodes, sphingolipid resolution stalls, phosphatidylserine asymmetry is restored inconsistently, and endothelial and barrier interfaces remain biochemically flagged as unresolved. These persistent signals sustain immune sentinel engagement by MAIT and γδ T cells without inflammation and impose a growing metabolic cost, reflected in nitrogen handling associated with prolonged repair. PEM therefore represents a recovery-phase control failure, not exertional injury, immune hyperactivation, or primary energy depletion.
Shear-exposed skeletal muscle
→ membrane coordination demand (PI/PLC turnover)
→ ER–mitochondrial Ca²⁺ timing failure at MAMs
→ recovery-phase ROS generation (not overload)
→ cardiolipin oxidation and imperfect remodeling
→ reduced mitochondrial recovery bandwidth
→ stalled sphingolipid resolution (ceramide → HexCer → SM)
→ intermittent failure to restore phosphatidylserine asymmetry
→ endothelial / barrier interfaces remain “not resolved”
→ MAIT & γδ T cells remain in sentinel mode
→ nitrogen cost accumulates as prolonged repair continues
Figure 3 — Recovery-phase failure stack (compressed)
Recovery logicInterpretive guardrail: This diagram illustrates timing and coordination during recovery, not causality or initiation. Signals shown represent recovery-phase persistence, not exertional injury, immune hyperactivation, or primary energy depletion.
This document integrates findings from AI-driven multi-omics modeling with established physiological, cellular, and lipid-biology principles to construct a mechanistic explanation of post-exertional malaise (PEM). The approach is interpretive and integrative, not experimental, and is intended to map reproducible biological state signatures onto a coherent recovery-phase control framework.
The primary empirical input is AI-derived multi-omics modeling that links post-exertional symptom severity to coordinated immune, metabolic, and lipid-associated signals. These data are treated as state markers—indicators of system behavior during PEM—rather than direct evidence of causal initiation. Additional support is drawn from established literature on microvascular physiology, ER–mitochondrial signaling, lipid remodeling, and immune sentinel biology.
This document focuses exclusively on PEM, rather than baseline disease features, because PEM represents the most mechanistically constrained and reproducible state in ME/CFS. Resting abnormalities are not assumed to reflect disease mechanism unless they scale with post-exertional worsening.
Recovery-phase failure, not exertional injury: PEM is modeled as a failure to terminate recovery processes after stress, rather than damage incurred during exertion itself.
Control precedes capacity: Upstream timing, coordination, and resolution mechanisms are treated as primary, with energetic or metabolic limitations emerging secondarily.
State dependence: Biological signals observed during PEM are assumed to reflect context-dependent system behavior, not fixed traits or constant pathology.
Immune signals as sentinels: Immune programs associated with PEM are interpreted as monitoring and containment responses to unresolved interfaces, not as primary inflammatory or autoimmune drivers.
Genetic contribution as modulation: Genetic variants are modeled as modifiers of recovery bandwidth, clearance efficiency, and tolerance to prolonged repair, rather than as disease-initiating lesions.
Observed associations (e.g., PI/PLC signaling, interferon-linked programs, lipid remodeling, nitrogen handling) are mapped according to known cellular roles and temporal constraints, with emphasis on signal termination, membrane resolution, and recovery bandwidth. Causality is assigned only where directionality is supported by established cell biology and physiological timing; otherwise, signals are treated as correlates of persistence.
This framework does not claim to identify a single root cause of ME/CFS, nor does it assert that all patients share identical mechanisms at baseline. It does not replace experimental validation and should be read as a mechanistically constrained synthesis that generates testable predictions, rather than definitive proof.
The purpose of this document is to provide a clear, internally consistent explanation of how PEM is generated and sustained across recovery cycles, using PEM as the organizing axis through which diverse biological findings can be coherently understood.
Xiong, R., Aiken, E., Caldwell, R., et al. (2025). AI-driven multi-omics modeling of myalgic encephalomyelitis/chronic fatigue syndrome. Nature Medicine, 31, 2991–3001. https://doi.org/10.1038/s41591-025-03788-3
The documents listed below define the conceptual and methodological framework used to interpret genetic signals and physiological mechanisms in this paper. Collectively, they establish layer boundaries, phenotype discipline, and phase dependence within the Gut–Liver–Autonomic (GLA) system architecture.
These materials are provided for transparency and interpretive context only. They are not cited as evidentiary sources and should be read as evolving systems-biology models used to organize and constrain interpretation, rather than as claims of mechanism or causation.
Core GLA framework documents
SMPDL3B phenotype frameworks
Control layers & system modulators