Santosh Manicka

Santosh Manicka

Affiliate Research Scientist · Allen Discovery Center, Tufts University

Understanding what biological systems compute and how

Interpretability Theoretical biology Bioelectricity Boolean networks Information theory Basal cognition Complex systems Morphogenesis

Non-neural biological systems routinely solve problems that look cognitive: a developing embryo coordinates millions of cells into a body plan, a gene regulatory network forms and recalls associative memories, an embryonic collective communicates to mount a rescue response under duress. We know a great deal about the molecular parts. What we lack is an understanding of what these systems are actually computing: what is represented, how information flows across space and time, where the decision boundaries lie, and how the outcome stays robust.

representations? information flow? boundaries? robustness?
The interpretability gap
opaque interpretable
Minimality as interpretability strategy

That is the problem I work on. My approach has two steps. First, I build deliberately minimal models, stripped to the fewest components that still produce the behavior of interest. Minimality is not an aesthetic choice; it is a strategy for understanding. A model with two essential variables can be fully understood and still yield deep insights; the same analysis becomes far more challenging as that number grows.

Second, I develop and apply analytical methods (spatiotemporal sensitivity fields, causal network decompositions, information-theoretic measures) that extract the computational and cognitive structure hidden in these models. In a 2019 framework paper (The Cognitive Lens), I laid out this program: treat biological patterning as information processing, then use integration methods to read out what the system represents, when it shifts from being externally driven to self-maintaining, and which cells integrate information from which others. The goal is not just to make models interpretable, but to discover whether cognition-like processes such as decision-making, memory, and communication are genuinely at work in non-neural biology.

Detailed Analysis Sifted
The Cognitive Lens framework

Each project builds a minimal model or a new measure,
then uses the cognitive lens to ask: what is this system computing?

Selected Publications

Full list on Google Scholar.

In preparation
How gene regulatory networks might use attention to remember: a computational exploration
N. A-Rahim, S. Manicka, M. Levin · Preprint in progress
Boundary Codes and Bulk Patterns: Towards Generative Holographic Morphogenesis
S. Manicka, M. Levin · Preprint in progress
2025
Field-mediated bioelectric basis of morphogenetic prepatterning
S. Manicka, M. Levin · Cell Reports Physical Science, 6(10)
2023
The nonlinearity of regulation in biological networks
S. Manicka, K. Johnson, M. Levin, D. Murrugarra · npj Systems Biology and Applications, 9, 10
2022
Effective connectivity determines the critical dynamics of biochemical networks
S. Manicka, M. Marques-Pita, L. M. Rocha · Journal of the Royal Society Interface, 19(186)
Biological underpinnings for lifelong learning machines
D. Kudithipudi, …, S. Manicka, …, H. T. Siegelmann · Nature Machine Intelligence, 4(3), 196–210
2021
Gene regulatory networks exhibit several kinds of memory
S. Biswas, S. Manicka, E. Hoel, M. Levin · iScience, 24(3), 102131
2019
Modeling somatic computation with non-neural bioelectric networks
S. Manicka, M. Levin · Scientific Reports, 9, 18612
2008 – 2017
The role of canalization in the spreading of perturbations in Boolean networks
S. Manicka · PhD Dissertation, Indiana University (2017)
Local ultrastability in a real system based on programmable springs
S. Manicka, E. A. Di Paolo · ECAL Proceedings 2009
'Psychoanalysis' of a minimal agent
S. Manicka, I. Harvey · ALIFE Proceedings 2008

Open-source research code

ElectricMorphogenesis

Bioelectric field models of tissue-level voltage patterning

NeuralPlatePatterning

Genetic-bioelectric model of embryonic frog brain patterning

BioElectricNetwork

Library for simulating, training, and evolving minimal bioelectric networks

MinimalDevelopmentalComputation

Hierarchical regenerative networks with local intracellular controllers

Boolion

Taylor series decomposition and continuous relaxation of Boolean networks

BooleanCanalization

R library for effective connectivity and unified canalization

PlatonicCriticality

Do the integers themselves exhibit scale-free criticality?

PsychoanalysisMinimalAgent

Evolving CTRNNs for the Secretary Problem: interpretability of a minimal agent (Artificial Life 2008)

ReferentialCommunication

Recurrent neural net agents evolving referential communication strategies (Artificial Life 2012)

UltrastableProgrammableSpring

Ashby-inspired adaptive control for a programmable spring actuator (ECAL 2009)

ChessSymbolicAI

Chess engine with minimax and alpha-beta pruning in Turbo C++ (2000)

2017
Ph.D. in Complex Systems
Indiana University Bloomington, U.S.A.
2008
M.Sc. in Evolutionary and Adaptive Systems
University of Sussex, U.K.
2003
B.E. in Computer Science and Engineering
College of Engineering, Guindy, India