Summary:
A mathematical and computational framework published in The Innovation resolves a foundational challenge in empirical science: determining genuine cause and effect in complex, non-linear systems without physical intervention. Developed by researchers from Fudan University, the Chinese Academy of Sciences, and the University of Tokyo, the Interventional Dynamical Causality (IntDC) framework uses a metric called Interventional Embedding Entropy (IEE) to infer intervention-level causal directionality solely from observational time-series data. By leveraging state-space delay embedding, the method eliminates spurious associations created by hidden common confounders, successfully reconstructing functional neural connectomes in C. elegans, epidemic transmission routes, and gene regulatory networks.
Key Facts:
- Intervention-Level Standard: Unlike classical methods (such as Granger causality, transfer entropy, and convergent cross mapping) that assess predictive dependence, IntDC tests whether an external perturbation on variable X would alter variable Y.
- Confounder-Free Delay Embedding: Leveraging Takens’ delay embedding theorem, the method shows that observational records of Y contain the mathematical manifold imprint of X’s perturbations, allowing Interventional Embedding Entropy (IEE) to reject spurious links induced by shared drivers (Z).
- Broad Biological Benchmarking: The algorithm accurately reconstructed whole-brain functional neural connectomes from C. elegans calcium imaging, mapped COVID-19 transmission routes in Japan, and delineated master regulatory genes in circadian transcriptional clocks.
Source: Fudan University / Chinese Academy of Sciences / University of Tokyo
Establishing true cause and effect represents the primary ambition of empirical science. Neuroscientists seek to identify the precise microcircuits that trigger motor actions; geneticists aim to map which master transcription factors dictate cell differentiation; and climatologists must determine which atmospheric forcings govern planetary warming.
Historically, the gold standard for verifying causality has required deliberate physical manipulation: intervene on an independent variable while holding all else constant, and observe whether the target outcome shifts.
Yet in the vast majority of complex real-world architectures, from chaotic climate systems to living human neural circuits, direct physical intervention is either technically unfeasible, financially prohibitive, or ethically impermissible. Consequently, researchers have had to rely on passive observational time-series data, using predictive tools like Granger causality, transfer entropy, or convergent cross mapping (CCM).
However, these classical methods measure “constructive dynamical causality”, identifying whether the historical trajectory of variable X aids in forecasting variable Y. When a third unmeasured factor (Z) simultaneously drives both variables, these methods often register false-positive causal arrows, mistaking synchronized confounding for true functional control.
Now, reporting in the journal The Innovation, an international research team led by Professor Wei Lin of Fudan University, Professor Luonan Chen of the Chinese Academy of Sciences, and Professor Kazuyuki Aihara of the University of Tokyo has unveiled a mathematical paradigm that extracts true, intervention-level causality from passive observational data alone.
Interventional Dynamical Causality (IntDC) and Embedding Entropy
The team’s framework introduces a formal distinction between observational dependence and interventional causality in dynamical systems. Under the stricter criteria of modern causal inference (such as Judea Pearl’s do-calculus), X can only be designated as a true cause of Y if an external intervention on X produces an observable alteration in Y.
To satisfy this criterion without physically manipulating the system, the researchers turned to delay-coordinate embedding theorems from non-linear dynamical systems theory. According to Takens’ theorem, the multi-dimensional manifold of an entire coupled dynamical system can be reconstructed from the lagged time-series history of a single observed variable.
The investigators deduced that if variable X exerts true interventional influence over variable Y, the reconstructed state-space manifold of Y inevitably preserves the topological imprint of X’s dynamic perturbations.
By defining a new information-theoretic metric, termed Interventional Embedding Entropy (IEE), the framework quantifies the directional flow of intervened information between state spaces. This allows algorithms to calculate causal strength directly from passive recordings without prior knowledge of the underlying differential equations and without inducing experimental perturbations.
Benchmarking Against Classic Confounder Traps
The research team tested the framework across standard computational stress tests, including coupled logistic maps, Hénon attractors, and chaotic recurrent neural networks.
In classic three-node configurations where a shared driver (Z) synchronously influences both X and Y without any direct interaction between them, traditional Granger causality and information-transfer tools systematically generate spurious X < Y connections.
IEE demonstrated complete resistance to these confounding drivers, correctly rejecting the false connections. Furthermore, the metric maintained high stability across variations in time-series length, measurement noise, embedding dimensions, and arbitrary coupling strengths.
From Nematode Connectomes to Circadian Clocks
The utility of the IntDC framework was validated across three distinct real-world biological and epidemiological domains:
- Whole-Brain Connectomics: Applying IEE to whole-brain calcium-imaging time series from the nematode Caenorhabditis elegans, the algorithm reconstructed functional neural connectomes, establishing directed causal signaling pathways between individual motor, sensory, and interneurons.
- Epidemic Transmission Modeling: Utilizing regional infection records from Japan’s COVID-19 pandemic waves, the method identified geographic transmission vectors and pinpointed epicenters driving secondary outbreak clusters.
- Circadian Gene Regulatory Networks: Analyzing high-throughput transcriptomic oscillations, the tool reconstructed gene-regulatory loops, isolating the core feedback mechanisms of master circadian pacemaker genes.
By translating passive time-series data into rankable, quantitative causal strengths, the IntDC/IEE framework provides a foundation for automated causal discovery across neuroscience, systems biology, and computational medicine, bridging observation and experimental validation.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context added by our staff.
About this neuroscience Research:
- Media Contact: Yong Han
- Source: Innovation Press
- Image Credit: Image credited to Neuroscience News
- Original Research is Open Access: The Innovation (Oct 1, 2026). “Deciphering interventional dynamical causality from non-intervention complex systems.” Authors: Jifan Shi, Yang Li, Juan Zhao, Siyang Leng, Rui Bao, Kazuyuki Aihara, Luonan Chen, and Wei Lin.
- DOI: 10.1016/j.xinn.2026.101358
Abstract
Deciphering interventional dynamical causality from non-intervention complex systems
Detecting and quantifying causal influence is a focal topic in the fields of science, engineering, and interdisciplinary studies. However, causal studies on non-intervention systems attract much attention but remain extremely challenging.
The delay-embedding technique provides a promising approach. In this study, we propose a framework named interventional dynamical causality (IntDC) in contrast to the traditional constructive dynamical causality (ConDC). ConDC, including Granger causality, transfer entropy, and convergence of cross-mapping, measures the causality by constructing a dynamical model without considering interventions.
A computational criterion, interventional embedding entropy (IEE), is proposed to measure causal strengths in an interventional manner. IEE is an intervened causal information flow but in the delay-embedding space.
Further, the IEE theoretically and numerically enables the deciphering of IntDC solely from observational (non-interventional) time-series data, without requiring any knowledge of dynamical models or real interventions in the considered system.
In particular, IEE can be applied to rank causal effects according to their importance and construct causal networks from data. We conducted numerical experiments on logistic dynamics, coupled-Hénon maps, and chaotic neural networks to demonstrate that IEE can find causal edges accurately, eliminate effects of confounding, and quantify causal strength robustly over traditional indices.
We also applied IEE to real-world tasks, including estimating neural connectomes of Caenorhabditis elegans, detecting COVID-19 transmission networks in Japan, and investigating regulatory networks surrounding key circadian genes. IEE performed as an accurate and robust tool for causal analyses solely from the observational data.
Both the IntDC framework and the IEE algorithm proposed in this study provide an approach to understanding causal influence in diverse non-intervention complex systems.
