Week 7 — IANNs, Fungal Materials & DNA Design

Part 1: Intracellular Artificial Neural Networks (IANNs)

1. Advantages of IANNs over Boolean Genetic Circuits

Traditional genetic circuits implement Boolean logic — AND, OR, NOT gates — where every output is binary: a gene is either on or off. IANNs offer three key advantages over this:

  1. Custom activation functions — Boolean logic is just a special case of an IANN. By choosing an activation function (e.g. a sharp step = Boolean; ReLU or tanh = analog), you can implement Boolean circuits as a subset or go beyond them entirely without redesigning the underlying architecture.
  2. Non-linearity — non-linear activation functions allow the network to separate input combinations that a linear system can’t distinguish, enabling richer input-output mappings from the same number of genetic components.
  3. Graded, continuous inputs — real biological signals (transcription factor concentrations, metabolite levels) exist on a spectrum, not just “present” or “absent.” IANNs integrate these natively without thresholding, preserving signal information that Boolean circuits discard.

2. Useful Application for an IANN

Application: thermoregulatory soft robotic wearable

A knitted, soft-robotic wearable that autonomously regulates body temperature using two biological sensors and an IANN to integrate them:

  • Input X1: sweat biosensor — detects sweat concentration (NaCl, lactate) as a continuous electrochemical signal
  • Input X2: mechanosensor — detects muscle vibration frequency as a proxy for shivering

The IANN hidden layer integrates both graded signals with learned weights. The key insight is that neither input alone is sufficient: a wearer can be slightly sweaty and slightly shivering simultaneously (e.g. after exercise in cold weather), and the correct actuator response is neither fully open nor fully contracted. A Boolean circuit would have to threshold each signal independently and lose that nuance; the IANN interpolates a response proportional to both.

  • Output: soft pneumatic or shape-memory actuators woven into the fabric — high sweat + no shiver → structure opens (ventilates, cools); low sweat + high shiver frequency → structure contracts (insulates, retains heat)

Limitations: protein-based signal transduction is orders of magnitude slower than electronic sensing; biological noise and signal crosstalk between the two inputs could cause erratic actuation; and the IANN weights would need to be tuned to individual wearers’ sweat profiles and shiver thresholds.

3. Diagram: Intracellular Multilayer Perceptron

The assignment provides a single-layer perceptron where:

  • X1 = DNA encoding Csy4 endoribonuclease
  • X2 = DNA encoding a fluorescent protein whose mRNA is regulated by Csy4

A multilayer perceptron adds a hidden layer between input and output. In the intracellular context, layer 1 would express an endoribonuclease that post-transcriptionally regulates layer 2 outputs.

Intracellular Multilayer Perceptron — Csy4 as hidden node Intracellular Multilayer Perceptron — Csy4 as hidden node

Part 2: Fungal Materials

1. Existing Fungal Materials

The most compelling recent work on fungal materials comes from Jasmine Lu, whose research uses living mycelium as an interactive interface — essentially a biological tamagotchi. The fungus responds to care (light, humidity, touch) and the project deliberately questions what kind of relationship users develop with a living material versus a digital one. It’s a provocation as much as a technology: if your interface is alive, do you feel responsible for it? This framing is particularly relevant for wearables and soft robotics, where the boundary between tool and organism starts to blur.

At the infrastructure level, mycelium has also been shown to transmit electrical signals along hyphal networks — Andrew Adamatzky’s work on “fungal computers” demonstrated measurable voltage spikes in response to chemical and physical stimuli, opening the door to mycelium as a biological sensing and signaling substrate.

On the commercial side, several fungal materials have reached market:

  • Mycelium composites (Ecovative) — mycelium grown on agricultural waste forms a lightweight, biodegradable foam used for packaging and insulation. Produces far less CO₂ than styrofoam; limitation is moisture sensitivity and lower mechanical strength.
  • Fungal leather (Bolt Threads Mylo) — mycelium-based leather alternative used by Stella McCartney. Sustainable alternative to animal hide; scaling and long-term durability remain challenges.
  • Mycoprotein / Quorn — Fusarium venenatum fermented into a high-protein meat substitute. Established at scale; some allergenic potential in sensitive individuals.
  • Chitin — extracted from fungal cell walls, used in wound dressings and sutures. Biodegradable, biocompatible, and naturally antimicrobial.
  • Mycoremediation — species like Pleurotus ostreatus (oyster mushroom) can degrade petroleum, heavy metals, and some plastics. Paul Stamets has championed this as a low-cost environmental remediation tool.
  • Koji / Tempeh — Aspergillus oryzae and Rhizopus species underpin miso, sake, and soy sauce fermentation, and are now being explored as direct protein sources in their own right.

2. Genetic Engineering of Fungi

A useful way to frame fungal synthetic biology is to contrast it with bacterial engineering. In a project like synthesising casein from E. coli, bacteria are the fabricator — they express and secrete the protein, but the material itself is what they produce, not what they are. Mycelium flips this: the organism is the material. You’re not harvesting what it makes, you’re growing the structure itself.

This distinction shapes what genetic engineering of fungi looks like. Rather than inserting a production pathway, you’d engineer the fungus’s own structural properties — chitin composition, hyphal branching density, surface chemistry — to tune the material directly. For example:

  • Mechanical tuning — alter chitin synthase expression to make mycelium stiffer or more flexible on demand, useful for structural composites or soft robotics
  • Electrical conductivity — engineer metalloprotein networks or conductive surface coatings into hyphae for embedded sensing
  • Growth control — tune branching patterns to direct self-assembly into specific geometries without a mold

Fungi also offer practical advantages over bacteria for material applications: they self-assemble into complex 3D networks with no bioreactor required; their eukaryotic machinery supports post-translational modifications (glycosylation, disulfide bonds) needed for mammalian proteins; they grow on agricultural waste with minimal inputs; and many species are GRAS, lowering the regulatory barrier for consumer products.

Part 3: DNA Twist Order

Final Project DNA Design

In progress.

DNA Twist order submitted. Sequences and vector design finalized and sent for synthesis.