BioBlobs: Aggregation & Emergence

Updated 14 July 2026

BioBlobs are heterogeneous, blob-like entities spanning biological, computational, and physical systems, defined by emergent collective properties.
Worm blobs form active, viscoelastic collectives with oxygen-dependent dynamics that leverage topological entanglement for non-equilibrium behavior.
Proteinoid microspheres and bacterial aggregates generate electrical spikes and fluidic dynamics, underpinning unconventional computing and biofilm modeling.

BioBlobs is a heterogeneous term used in recent arXiv literature for several classes of blob-like biological, life-like, and computational entities whose salient behavior arises from aggregation, entanglement, self-assembly, or adaptive partitioning. In one line of work, BioBlobs are worm collectives formed by long, slender aquatic worms such as Lumbriculus variegatus and Tubifex tubifex, which behave as active viscoelastic matter and exhibit oxygen-dependent mechanics (Deblais et al., 2023, Tuazon et al., 2022). In another, BioBlobs are proteinoid microsphere networks that generate endogenous electrical spikes, respond to illumination, and are studied as substrates for unconventional computing (Adamatzky, 2021, Mougkogiannis et al., 2023). Related work treats bacterial aggregates in microflows in BioBlob terms (Carpio et al., 2024), while a separate protein representation learning literature uses “BioBlobs” for a differentiable graph-partitioning module that identifies function-relevant protein substructures (Wang et al., 2 Oct 2025).

1. Scope and terminology

Across the cited literature, “BioBlobs” does not denote a single standardized object class. Instead, it names several systems unified by blob-like organization and emergent collective properties. This suggests that BioBlobs is currently best understood as a cross-domain descriptor rather than a fixed taxonomic term.

Context	System	Defining use
Entangled living matter	Worm blobs	Active viscoelastic, three-dimensional soft entity with solid–liquid duality
Unconventional computing	Proteinoid microspheres	Photosensitive, electrically spiking microsphere networks with programmable growth
Microflow aggregation	Bacterial aggregates	Biofilm growth, erosion, motion, and streamer formation in laminar microflows
Protein representation learning	BioBlobs module	Differentiable partitioning into flexibly-sized, non-overlapping substructures

The term’s polysemy is not merely semantic. In the worm and proteinoid literatures, BioBlobs refers to materially instantiated collectives with emergent mechanics or excitability. In the protein representation learning literature, the same name denotes an architectural module that abstracts coherent substructures from protein graphs. The shared conceptual emphasis is on coherent subunits whose functional properties are not well captured by isolated constituents or rigid fixed-size decompositions (Deblais et al., 2023, Adamatzky, 2021, Wang et al., 2 Oct 2025).

2. Worm blobs as topological active matter

Worm blobs are described as aggregates formed by aquatic worms such as Lumbriculus variegatus and Tubifex tubifex. These aggregates are physically and topologically entangled, and they exhibit solid–liquid duality: they can behave as soft solids over short timescales and flow like liquids over longer timescales. The blob is characterized as an active viscoelastic, three-dimensional soft entity that can respond to temperature, light, oxygen concentration, and chemical gradients, and can break symmetry to locomote collectively through positional differentiation between pulling and friction-reducing worms (Deblais et al., 2023).

The entanglement is treated as a topological resource rather than a nuisance variable. The cited work frames worm blobs as “entangled living polymers” and “topological active matter,” emphasizing configurational trapping, effective cohesion, and mechanical stability despite repulsive interactions at the monomer level. Internal ultrasound imaging is used to construct a tangle graph in which edges encode intertwined pairs, leveraging the linking number from knot theory. A control parameter for tangling is the chirality number,

$\gamma = \frac{\alpha}{2\pi\lambda},$

where $\alpha$ is the rate of head turning and $\lambda$ is the turning reversal rate; higher $\gamma$ drives transitions from untangled to tangled states in simulation (Deblais et al., 2023).

Several observations distinguish worm blobs from passive polymer systems. Blob diffusion is reported to be independent of blob size, suggesting that only worms at the surface contribute to motion. Activity also modifies rheology in a nonclassical way: the zero-shear viscosity decreases with increased worm activity, opposite to trends in conventional polymer solutions. Coalescence, fragmentation, reversible transitions between flattened “pancakes” and spherical aggregates, and active responses to environmental gradients further motivate the interpretation of worm blobs as a living model system for non-equilibrium polymer physics, morphological computation, and bioinspired tangling soft robot collectives (Deblais et al., 2023).

3. Oxygenation-controlled collective dynamics

For Lumbriculus variegatus, oxygenation is a primary control variable for collective morphology, activity, and internal stress. In high dissolved oxygen, defined as $>8.0$ mg/L, blobs form tightly packed, spherical structures; individuals remain mostly immobile and strongly entangled. In low dissolved oxygen, defined as $<2.0$ mg/L, blobs become more dynamic and less compact, worms disentangle, and individuals actively extend and wave their tails upward in “tail reaching” to supplement respiration (Tuazon et al., 2022).

The experimental program combines a closed-loop respirometer with adjustable flow, flow-rate manipulation to levitate blobs and measure projected area, and physical lifting with a serrated endpiece attached to a linear actuator to probe internal stress. Three quantitative observables are central: relative activity flux in the tail-reaching region, exposed surface area in suspended blobs, and center-of-mass lift height in the lifting experiment (Tuazon et al., 2022).

Metric	Low DO	High DO
Mean relative activity flux	$30.6 \pm 8.5\%$	$0.4 \pm 0.6\%$
Mean exposed area	$298.9 \pm 16.3\ \text{mm}^2$	$217.1 \pm 13.6\ \text{mm}^2$
Mean $\alpha$ 0 in lifting	$\alpha$ 1 mm	$\alpha$ 2 mm

These measurements imply an approximately $\alpha$ 3 increase in relative tail-reaching activity flux in low DO and an approximately $\alpha$ 4 increase in exposed area under low DO. Under a higher flow rate of $\alpha$ 5 mL/s, the larger projected area in low DO indicates a more open configuration. In the lifting assay, high-DO blobs can be lifted off the bottom as a unit, whereas low-DO blobs fail to hold together, indicating that only highly entangled, high-internal-stress blobs are manipulable as a unit (Tuazon et al., 2022).

The cited interpretation is that low oxygen drives a transition toward a more open, dynamic, fluid-like state that increases respiratory efficiency through greater exposed area and tail lifting, while high oxygen promotes a dense, mechanically coupled, more solid-like state. The paper further suggests that the high-DO contraction may help prevent displacement by strong flows after rain or wind when oxygen is high. More broadly, the oxygenation results are presented as a basis for modeling active entanglement in swarm robots, self-assembly structures, and soft material entanglements (Tuazon et al., 2022).

4. Proteinoid microspheres as computational BioBlobs

Proteinoids are thermal proteins produced by heating amino acids to their melting point and initiating polymerization to produce polymeric chains. In aqueous solution, they swell into hollow microspheres and can form networked architectures connected by pores and tube-like outgrowths. The literature describes programmable growth through amino acid composition and thermal regime, with microsphere diameters of $\alpha$ 6– $\alpha$ 7 in one study, and composition-specific SEM morphologies including nanospheres and nano-needles in another (Adamatzky, 2021, Mougkogiannis et al., 2023).

These microspheres exhibit endogenous electrical activity. Reported steady-state membrane potentials are $\alpha$ 8– $\alpha$ 9 mV even without external stimulation, and action potential-like electrical spikes and oscillatory activities are routinely recorded. Spontaneous bursts can persist for days to weeks in stable aqueous solution, and activity amplitude and regularity depend on composition; the addition of lecithin is reported to increase amplitude (Adamatzky, 2021). The systems are also photosensitive: illumination induces a photovoltaic effect and alters electrical activity, while darkness slows and can cease discharges, and intense light can arrest oscillations at a higher membrane potential (Adamatzky, 2021).

A detailed photostimulation study resolves composition- and regime-specific spiking statistics. Under periodic illumination, the reported mean amplitude and mean period are $\lambda$ 0 mV and $\lambda$ 1 h for L-Phe:L-Lys, $\lambda$ 2 mV and $\lambda$ 3 h for L-Glu:L-Phe, $\lambda$ 4 mV and $\lambda$ 5 h for L-Phe, and $\lambda$ 6 mV and $\lambda$ 7 h for L-Glu:L-Phe:L-His. Under constant illumination, amplitude often decreases while frequency increases; for L-Glu:L-Phe, three spike zones are reported, including a large-spike zone with mean $\lambda$ 8 mV. No oscillations are observed with light off. The amplitude–intensity relation is summarized by the linear fit

$\lambda$ 9

with a statistically significant slope reported as $\gamma$ 0, $\gamma$ 1 (Mougkogiannis et al., 2023).

In the unconventional computing literature, proteinoid BioBlobs are treated as a medium of coupled oscillators. The computational analogy is made explicitly to Belousov–Zhabotinsky systems, reservoir computing, and in-materio computing. A simulation of excitation spreading in a layer of microspheres uses the FitzHugh–Nagumo equations, and electrode responses are interpreted as logical operations including OR, XOR, AND, and selector gates. Large ensembles are described as capable of implementing mappings $\gamma$ 2, thereby functioning as finite-state machines with substrate-defined transfer functions (Adamatzky, 2021). Within this usage, BioBlobs are self-assembling, living, or life-like aggregates whose information processing emerges from spatiotemporal excitation patterns and mutable connectivity rather than rigid circuits (Adamatzky, 2021).

5. Bacterial aggregates in microflows

A further BioBlob-relevant domain concerns bacterial aggregates and biofilms in laminar microflows. The cited work formulates a hybrid computational model in which biomass dynamics follow stochastic rules for adhesion, erosion, and motion, while the surrounding flow is updated through numerical approximations of the incompressible Navier–Stokes equations. The computational region is tiled at bacterial scale, $\gamma$ 3– $\gamma$ 4, and partitioned into fluid $\gamma$ 5, biofilm $\gamma$ 6, and interface $\gamma$ 7 (Carpio et al., 2024).

Two adhesion channels are modeled: attachment to surfaces, denoted $\gamma$ 8, and adhesion at the fluid/biofilm interface, denoted $\gamma$ 9. Erosion is shear-induced, with probability

$>8.0$ 0

where $>8.0$ 1 is the local shear force and $>8.0$ 2 is biofilm cohesion. Motion of boundary biomass tiles is treated analogously, for example

$>8.0$ 3

where $>8.0$ 4 is the force component in the $>8.0$ 5 direction. This construction couples interface evolution directly to local hydrodynamic stresses and cohesive strength (Carpio et al., 2024).

The model is used to study streamer formation in three-dimensional corner flows. Corners and regions with secondary flows or vortices are reported as preferential nucleation sites and as attractors for streamer initiation. Morphology depends on the balance among $>8.0$ 6, $>8.0$ 7, cohesion $>8.0$ 8, and the local shear and normal forces. Low $>8.0$ 9 yields weak attachment and premature detachment; low $<2.0$ 0 produces thin, fragile threads; higher $<2.0$ 1 or $<2.0$ 2 supports wider, more robust streamers able to bridge between corners; excessively large $<2.0$ 3 produces “blobby” spherical expansion rather than filaments (Carpio et al., 2024).

The control implications are practical. Thin homogeneous biofilms may serve as biosensors in MEMS without disrupting the microsystem, whereas uncontrolled streamer growth increases drag and may clog channels. The model therefore motivates design strategies based on channel geometry, shear control, and tuning of cohesion or adhesion to promote homogeneous films or suppress disruptive aggregate morphologies (Carpio et al., 2024).

6. Computational abstractions and representational extensions

The blob concept also appears in computational and mesoscale modeling frameworks that are not themselves biological aggregates but are relevant to BioBlob analysis and abstraction. One example is a real-time algorithm for detecting and tracking blob-filaments in fusion plasma. The method decomposes the task into local identification of feature cells, grouping feature cells into extended features by connected component labeling on triangular meshes, and frame-to-frame tracking through spatial overlap and centroid heuristics. On a 30 GB fusion simulation dataset, the implementation shows linear speedup on 1024 processes and completes blob detection in less than three milliseconds on Edison, a Cray XC30 system at NERSC (Wu et al., 2015). The paper explicitly notes that similar spatio-temporal features matter in combustion and medical imaging, indicating that “blob” has become a transferable analytical category.

A second example is the multiblob hydrodynamics framework, which represents a particle as a small collection of overlapping Peskin immersed-boundary kernels held together by springs. The principal construction uses 12 blobs at the vertices of an icosahedron, sometimes with an additional central blob. Although the surface is not explicitly resolved, the model reproduces translational and rotational mobilities, near-field flow, and a divergent lubrication force at finite inter-particle distance. It supports particles with radii in the range $<2.0$ 4, uses GPU acceleration in the Fluam code, and is presented as suitable for dense suspensions, complex-shaped particles, and soft or elastic objects, including BioBlobs in the broad soft-matter sense (Vázquez-Quesada et al., 2014).

A distinct semantic extension appears in protein representation learning, where “BioBlobs” names a plug-and-play, fully differentiable module for dynamically partitioning protein structures into flexibly-sized, non-overlapping substructures and quantizing them into an interpretable codebook. The system is integrated with a GVP-GNN backbone and includes a neural blob partitioner, blob codebook, and global–blob attention fusion. Reported results include improvements over standard baselines; for Enzyme Class under a structure split, the scores reported are 0.621 for GNN, 0.451 for GVP-GNN, and 0.684 for BioBlobs, while for Gene Ontology under a random split, 0.569 is reported for GVP-GNN and 0.817 for BioBlobs. The learned blobs are described as corresponding to known motifs such as $<2.0$ 5-helices, $<2.0$ 6-sheets, and catalytic sites, and performance is reported to peak with a moderate number of approximately 5 blobs of size 10–15 residues (Wang et al., 2 Oct 2025).

Taken together, these extensions show that BioBlob-related thinking now spans materially embodied collectives, hydrodynamic surrogates, spatiotemporal detection pipelines, and representation-learning architectures. A plausible implication is that the term’s scientific utility lies less in denoting a single material substrate than in identifying coherent, functionally consequential substructures whose behavior emerges from coupling, context, and collective organization.