Notes

Reality is not made of things, but of possibilities. We do not create ex nihilo; we transform latent possibilities into coherent structures. When coherence reaches a critical threshold, a new level of ontology emerges.

Notes

About these Notes

This page collects thoughts, working hypotheses, conceptual sketches, and open questions that emerge during research. They are not intended to be definitive statements, but rather snapshots of ideas as they evolve.

A Home for an Idea: Toward a Mesoscopic View of Biology

Lunan Foldomics was not created merely as a company, nor as an attempt to imitate the structure of an academic laboratory or private research institute. It was created to provide a rigorous intellectual home for an idea.

The starting point is a particular view of independent research. An independent scientist should not be someone who stands outside the scientific community, but someone who willingly submits ideas to its highest standards: peer review, reproducibility, methodological transparency, open criticism, and continuous revision. Independence, in this sense, is not isolation. It is the construction of an autonomous environment in which a research program can mature while remaining fully accountable to the scientific method.

The central scientific motivation arises from a simple observation. Modern biology has achieved extraordinary descriptive depth. Genomics, transcriptomics, proteomics, metabolomics, spatial biology, and digital pathology allow us to characterize biological systems with unprecedented precision. Yet, paradoxically, as these descriptive layers become increasingly detailed, they also become increasingly disconnected. Contemporary biology faces not only an abundance of data, but the challenge of organizing heterogeneous information into coherent representations.

This observation motivates what we tentatively call a mesoscopic view of biology.

The hypothesis is not that there exists a new biological object waiting to be discovered between genes and tissues. Rather, there may exist a new representational space in which relationships between different biological descriptions become observable, navigable, compressible, and ultimately predictive. This mesoscopic space is not defined by molecular components, but by the organization of information itself.

Within this framework, artificial intelligence becomes more than a computational tool. It becomes a possible mesoscope.

When a neural network robustly predicts gene expression from tissue morphology—or vice versa—it may not be merely discovering superficial correlations. It is constructing an internal representation that belongs neither to morphology nor to transcriptomics alone. Instead, it generates a latent relational structure through which these domains become mutually translatable.

This immediately raises a fundamental question.

Are these latent spaces themselves biological entities, or are they simply sophisticated mathematical constructions?

Our current position does not attempt to eliminate this tension. Instead, it reframes it. Perhaps latent spaces should not be understood as biological objects at all. Perhaps they are better understood as epistemological observables—interfaces that emerge whenever different descriptive layers become mutually intelligible.

This perspective naturally touches philosophical questions.

The mesoscopic space is not a molecule, an organelle, or a physical structure that can be directly measured. It is a condition of representation. Like Kant's metaphorical "glove" separating the observer from reality, it does not belong entirely to either the observed system or the observer. Rather, it belongs to the act of organized observation itself. Wet biology remains indispensable because it provides both the source and the constraint of biological knowledge. However, the primary object of investigation shifts toward understanding how biological information is organized, compressed, transferred, and represented across multiple descriptive scales.

From this perspective, Evoscope should not be understood as an isolated project, but as one manifestation of a broader research program.

Evoscope was conceived as a synthetic environment in which simple regulatory rules generate emergent morphologies under controlled conditions, allowing us to ask whether shared latent representations can arise between regulatory programs and spatial organization. Although intentionally minimal, Evoscope addresses a larger question:

What representations make biological organization observable when regulatory, morphological, and functional descriptions are not directly translatable into one another?

For this reason, Lunan Foldomics should never aspire to become the exclusive home of mesoscopic biology.

Its greatest success would be precisely the opposite.

If the concepts developed here become useful, they should leave Lunan Foldomics, be criticized, improved, extended, and eventually adopted by the broader scientific community. Like the strongest scientific ideas, their ultimate purpose is not to remain proprietary, but to become part of the ordinary language of science—to the point where they are eventually used without conscious reference to their origin.

The institutional philosophy follows naturally from this vision.

Lunan Foldomics is not intended to function primarily as a commercial startup, nor as a closed intellectual school centered around its founder. It is conceived as a lightweight, computationally driven, open scientific organization devoted to producing rigorous knowledge. Future consulting activities, collaborations, software implementations, educational initiatives, or commercial applications may eventually support the research program financially, but they should never become its primary purpose.

The primary mission remains unchanged:

to preserve, develop, and experimentally challenge the hypothesis that biology requires new computational observables capable of connecting descriptive layers that are currently treated as independent domains.

The ambition, therefore, is not to claim the invention of a "new kind of science." It is to propose a new way of observing biological systems.

Just as theoretical physics developed mathematical languages that generated experimentally testable predictions, mesoscopic biology seeks to develop computational languages capable of describing relationships across biological scales. Experiments remain the ultimate arbiters of scientific validity. Theory provides the representations that make new experiments conceivable.

Lunan Foldomics should therefore pursue its research with maximum ambition—as if mesoscopic biology could become an important component of the future of biological science—while maintaining maximum scientific discipline, recognizing that every concept must ultimately survive through evidence, prediction, reproducibility, and critical evaluation by the scientific community.

If successful, the greatest achievement of this research program will not be that Lunan Foldomics becomes indispensable.

It will be that the questions it asks become impossible to ignore.

Form as the Signature of Biology

Research Note #003 · July 1st, 2026

Morphology is not merely what biology shows. It is often the first way in which biology becomes observable, discussable, and ultimately understandable.

A biological form is not an ornament of life. It is not simply the aesthetic consequence of molecular processes unfolding beneath the surface. It is a record. A trace. The point at which genetic regulation, tissue mechanics, cell adhesion, proliferation, cell death, signaling, and the temporal history of a living system converge into something that can be observed.

Form is a solution.

This is what makes morphology so unique among biological observables. When we measure gene expression, proteomics, or metabolomics, we obtain vectors of numbers. Their organization must be reconstructed, and often much of their spatial context has already been lost. Morphology, instead, is born organized. It already contains relationships, boundaries, gradients, interfaces, symmetries, asymmetries, densities, and spatial constraints.

In solid tumors, this becomes immediately apparent. An invasive front, a metastatic lesion, a glandular architecture, or a necrotic region are not simply accumulations of cells. They are organized configurations. They behave almost like pathological sub-organs, growing, moving, and occupying space according to their own internal procedural logic.

Liquid tumors appear to challenge this perspective. Leukemias, for example, possess no obvious mass, no architectural boundary, no readily identifiable geometry. They resemble clouds of cells moving throughout the body rather than localized structures. Yet even a swarm has a shape.

The sea itself has no permanent contour, and yet we immediately recognize its morphology. A fluid, a gas, or a dispersed cellular population can still possess an observable organization, not through rigid boundaries but through distributions, densities, trajectories, collective dynamics, and energetic states. Morphology, therefore, should not be reduced to static geometry. It may also describe dynamic configurations unfolding across space and time.

This observation carries an important consequence. Morphology does not contain everything.

No single observable can fully capture the complexity of a biological system. Thermodynamic variables, metabolic states, mechanical forces, and countless molecular processes cannot always be directly inferred from form alone. But this is not the question.

The real question is different.

How much regulatory information remains embedded within biological form?

This is an experimental question. If morphology preserves even a structured fraction of a system's internal regulatory state, then it becomes possible to interrogate it. One may ask whether a cellular configuration contains sufficient information to recover latent regulatory programs, identify gene networks, infer developmental trajectories, or predict underlying biological states.

In this sense, morphology is the signature of biology. Not a static identity card, but a signature. A signature carries history. It reflects constraints, variations, decisions, and the path that generated it. Two signatures belonging to the same individual are never perfectly identical, yet both unmistakably preserve the identity of their author. Biological forms behave in much the same way. They are not exact replicas of one another, but they preserve traces of the generative processes that produced them.

This perspective naturally extends beyond morphology itself. The broader question is not whether morphology is special, but whether living systems continuously leave structured traces of their internal organization within their observable states. Morphology may simply be the first and most intuitive observable. Tomorrow, that observable might instead be a spatial transcriptomic field, a metabolic landscape, a mechanical tissue configuration, or the collective dynamics of a moving cell population. The fundamental question remains unchanged:

How much of a living system's internal state can be recovered from its observables?

Evoscope was conceived within this question. Not as a complete model of morphogenesis, but as a minimal experimental framework designed to test whether biological form contains partial yet structured information about the regulatory processes that generated it.

If this hypothesis proves correct, morphology is more than the visible appearance of biology.

It is one of its enduring memories.

Towards a Mesoscopic Representation Principle

Research Note #002 · July 1st, 2026

Motivation

The Mesoscopic Representation Principle did not emerge in isolation. Rather, it represents an attempt to unify several established ideas from machine learning, theoretical biology, and complex systems into a common conceptual framework.

The principle draws inspiration from four major research directions:

Representation Learning

Manifold Hypothesis

Mesoscopic Biology

Morphogenetic Landscapes

1. Representation Learning

Modern representation learning suggests that high-dimensional observations can often be compressed into compact latent spaces while preserving the information necessary for prediction.

Relevant concepts include:

Information Bottleneck
Sufficient representations
Variational Autoencoders
Disentangled representation learning

These methods motivate the idea that a latent space can preserve biologically meaningful information.

2. The Manifold Hypothesis

The manifold hypothesis proposes that high-dimensional biological observations actually occupy a much lower-dimensional manifold.

This idea has become central in:

single-cell transcriptomics
developmental trajectories
morphometrics
generative biology

The Mesoscopic Representation Principle extends this concept by proposing that the manifold may simultaneously encode multiple biological scales.

3. Mesoscopic Biology

The notion of a mesoscopic scale is well established in

statistical physics
soft condensed matter
tissue biology
complex systems

However, here the term mesoscopic is used differently.

Rather than indicating only an intermediate physical scale, it denotes a latent state space that mediates information between molecular regulation and emergent phenotype.

This operational definition distinguishes the present framework from traditional mesoscale descriptions.

4. Morphogenetic Landscapes

From Waddington's epigenetic landscape to gene regulatory attractors and developmental trajectories, biology has long sought a bridge between molecular regulation and morphology. The Mesoscopic Representation Principle proposes that this bridge may be understood as a compact latent representation discoverable through modern artificial intelligence.

Then...

Mesoscopic Representation Principle

"Whenever two distant biological scales admit a common latent representation that is compact, learnable, and navigable, this representation defines a mesoscopic state space containing the information necessary to connect molecular regulation with macroscopic phenotype."

What is new?

The novelty of this proposal does not lie in any individual concept listed above. Rather, it lies in the hypothesis that these ideas are manifestations of a common principle:

"A navigable latent space simultaneously connecting molecular regulation and morphology constitutes an operational definition of a biological mesoscopic state."

From Principle to Axiom...

The previous principle naturally suggests a stronger formulation. Although still speculative, one may ask whether the existence of a common latent representation could be elevated from an empirical observation to a foundational assumption of mesoscopic biology.

Proposed Mesoscopic Representation Axiom

"Whenever two observable biological scales are jointly reconstructable from the same compact, learnable, and navigable latent state, that latent state defines the minimal mesoscopic representation of the biological system."

Why an axiom?

If accepted, this statement changes the role of latent representations. Rather than being considered merely computational embeddings, they become fundamental biological state spaces from which multiple observable scales emerge as different projections of the same underlying organization.

Core inspirations

Waddington, C.H. — The Strategy of the Genes (1957)

Tishby, Pereira & Bialek — “The Information Bottleneck Method” (1999/2000)

Bengio, Courville & Vincent — “Representation Learning: A Review and New Perspectives” (2013)

Goodfellow, Bengio & Courville — Deep Learning (2016)

This note builds on ideas from Waddington’s morphogenetic landscape, the information bottleneck principle, representation learning, and the manifold hypothesis, but reframes them into an operational definition of a mesoscopic biological state space.

The Latent Language of Morphogenesis

Reading the Evoscope Autoencoder as a Mesoscopic Generator

Research Note #001 · June 2026

A conceptual reading of the Evoscope autoencoder. This note explores how latent representations may be interpreted as mesoscopic descriptions linking morphology and regulation.

Introduction

This note emerged from the development of Evoscope, a computational framework designed to study how multicellular morphologies relate to underlying regulatory states (see the Research page for project details and associated publications). In Evoscope, synthetic tissues are generated by populations of interacting cells carrying simplified regulatory programs ("genomoids"), and a convolutional autoencoder is trained to learn compact representations of these emergent morphologies while simultaneously predicting their regulatory states.

The present note is not a technical description of Evoscope itself. Rather, it explores a conceptual question raised by the model: what kind of information is actually stored within the latent representations that connect morphology and regulation?

Figure 1. Representative Evoscope morphology. Example of a synthetic multicellular morphology generated by Evoscope 0.9.1. While the underlying genomoid regulatory states are not directly visible, they contribute to the emergence of the spatial organization observed here. The latent representations discussed in this note can be viewed as mesoscopic descriptions linking these hidden regulatory processes to their observable morphological outcomes

Autoencoders are commonly described as dimensionality reduction algorithms. Their purpose is usually presented in practical terms: compress a high-dimensional input into a smaller latent representation and then reconstruct the original data as accurately as possible.

The Evoscope v0.9.1 autoencoder can certainly be understood in this way. Given a morphology snapshot represented as a one-hot encoded spatial grid, the model learns a compressed latent representation capable of reconstructing the original morphology.

However, such a description only captures part of what is happening.

When examined in the context of Evoscope and Foldomics, the autoencoder can be interpreted as a mechanism that searches for a mesoscopic description of a biological system: a representation that lies between morphology and regulation.

The architecture suggests a simple but powerful question: can a compressed representation of morphology become informative enough to reconstruct not only form, but also the underlying regulatory state? This question motivates the interpretation developed in the following pages.

The Encoder: From Morphology to Latent Description

The encoder receives a morphology snapshot as input. Through successive convolutional layers, spatial resolution is progressively reduced while the number of feature channels increases. Conceptually, the transformation can be represented as:

Morphology Snapshot

↓

32 feature maps

↓

64 feature maps

↓

128 feature maps

↓

Latent Space z

The encoder therefore performs two operations simultaneously:

First, it reduces the dimensionality of the original morphology.
Second, it constructs increasingly abstract descriptions of the observed spatial organization.

In this view, the encoder is not merely compressing information. It is generating candidate descriptions of the observed system. The resulting latent vector z can be interpreted as a compressed hypothesis about the morphology.

The Latent Space as Functional Compression

The latent space is often described as a compressed version of the input. In Evoscope, however, this interpretation is incomplete.

The latent coordinates are not required only to reconstruct morphology. They must also support the prediction of regulatory variables through a second branch of the network.

Consequently, z is not simply a visual compression:

It becomes a functional compression.
Information that is useful only for visual reconstruction may disappear.
Information that is useful for predicting regulatory state may be preserved.

The latent space therefore evolves toward a representation that balances multiple constraints simultaneously. This is the origin of its mesoscopic character.

Figure 2. Annotated reading of the Evoscope v0.9.1 autoencoder architecture.

Decoder and Gene Head

Two independent observers receive the same latent representation.

The Decoder attempts to reconstruct morphology.
The Gene Head attempts to predict regulatory variables.

Both receive exactly the same input: z. Yet they ask different questions.

The Decoder asks:

"Does z contain enough information to reconstruct the original form?"

The Gene Head asks:

"Does z contain enough information to infer the regulatory state?"

The training objective combines both requirements:

Loss = α · Loss_reconstruction + β · Loss_gene

The latent space is therefore shaped by two simultaneous pressures:

preservation of morphology
predictability of regulation

Neither objective dominates completely.

The final representation emerges as a compromise between them.

Emergence of Mesoscopic Coordinates

An important observation follows from the architecture. Regulatory variables are not directly used as input to the encoder.

The encoder only observes morphology. Nevertheless, the gene-prediction loss propagates backward through the network and modifies the encoder itself.

As training progresses, the encoder is encouraged to preserve those morphological features that are predictive of regulatory state. The latent coordinates therefore become progressively enriched with information that links morphology and regulation.

In this sense, the mesoscopic coordinates are not predefined. They emerge from the negotiation between reconstruction and prediction.

The Poet, the Sculptor, and the Painter

A useful metaphor can help illustrate this process. Imagine three artists. The Encoder is a poet. The Decoder is a sculptor. The Gene Head is a painter. The poet observes a group of models. He studies their shape, posture, appearance, and all visible characteristics.

Instead of creating a sculpture or a painting, he writes a short poem. The poem is compressed. It cannot contain every detail. It must contain only what is essential.

This poem is the latent representation z. The poem is then sent to the sculptor and the painter. The sculptor has never seen the original models.

He reads the poem and attempts to reconstruct their form.

The painter has never seen the original models either. She reads the same poem and attempts to reconstruct their colors and defining traits.

The poet then compares their work to the original models. If important information has been lost, he rewrites the poem.

The process repeats. Gradually, the poem becomes capable of evoking both form and regulation.

The latent space is therefore neither a sculpture nor a painting. It is a language.

A compact description from which multiple aspects of reality can be reconstructed.

Figure 3. The Poet, the Sculptor and the Painter: a metaphorical interpretation of latent-space learning.

Conclusion

From a conventional machine-learning perspective, the Evoscope autoencoder is a convolutional autoencoder equipped with a regulatory prediction head. From a mesoscopic perspective, it can be viewed differently. The encoder becomes a generator of compressed descriptions.

The latent space becomes a language rather than a storage container. The decoder and gene head become independent interpreters of that language. In this interpretation, z is neither morphology nor regulation. It is a mesoscopic description capable of evoking both.

The autoencoder therefore becomes more than a dimensionality reduction algorithm.

It becomes a generator of mesoscopic representations.

Credits

Author: Luca Zammataro

Lunan Foldomics LLC

Prepared with the assistance of generative AI tools for drafting, illustration, and editorial support.

Page updated

Google Sites

Report abuse