What We Do How It Works DigitalSoma Digital Pedon Projects About Get In Touch
Biological & Ecological Digital Twins
BioTwinR̂ — Biological Digital Twins

Digital twins for
living systems

BioTwinRs builds model-based digital twins for biological and ecological systems — computational replicas that mirror real processes in real time, enable you to predict, optimise, and understand without guesswork. Our flagship framework, DigitalSoma, brings this to animal physiology: any taxa, any sensor, from herd welfare to regulatory pharmacovigilance.

What We Do

Digital twins for living systems

A biological digital twin is a computational object that mirrors a living system in real time — continuously updated from sensor data, executing model (mechanistic or data-driven), and able to infer variables that are difficult or impossible to measure directly. We apply this paradigm across animals, organisms, and whole agroecosystems.

🦋
Animal physiology twins

DigitalSoma represents any animal — livestock, companion, wildlife, or aquatic — as a continuously updated physiological digital twin. Sensor streams from wearables, implants, and lab assays feed a composable solver chain that infers cardiac output, metabolic rate, thermal stress, and oxygen consumption in real time.

📊
Herd & population intelligence

Scale from individual animals to herd-level or population-level monitoring. Digital twins for cattle, sheep, and aquaculture flocks track welfare continuously, flag health events early, and feed regulatory pharmacovigilance pipelines automatically — without manual data entry.

🧬
In-silico biological models

Physics(Physiology)-informed computational models that replicate organ and whole-body responses to drugs, toxins, and environmental stressors — reducing the need for animal experimentation in line with EU and FDA New Approach Methodologies and the 3Rs framework.

🌿
Agroecosystem & soil twins

From a single soil horizon to a full farm system — our agroecosystem twins (e.g., SoilPedon) fuse real-time sensor data with soil physics to support precision agriculture, carbon accounting, and integrated animal–environment monitoring at landscape scale.

How It Works

From sensor to foresight

Every BioTwinR̂ platform follows the same four-step process, regardless of the biological domain — animal, organism, or agroecosystem.

01
Ingest any observational stream

Any sensor, instrument, or laboratory assay can feed the twin through a schema-agnostic six-field manifest contract. Wearable biosensors, implanted devices, remote thermal imaging, blood panels, environmental monitors — all are ingested through the same interface. Units are converted automatically; no preprocessing or schema changes are required.

02
Run a composable solver chain

Every time new data arrives, the twin executes a directed acyclic chain of physics-based and process-based solvers — not statistical models. For animal twins this means cardiovascular, metabolic, thermoregulatory, and respiratory equations running in sequence. Each solver consumes the outputs of those before it, so high-order variables like cardiac output, thermal comfort index, and physiological stress index are inferred automatically without redundant computation.

03
Simulate scenarios before they happen

With a calibrated twin running, you can project forward: how will this animal’s physiological state evolve under heat stress over the next six hours? What is the expected drug response trajectory given this animal’s current metabolic baseline? The twin answers in milliseconds, without waiting for observable symptoms to appear.

04
Act on alerts and export to any system

Threshold alarms fire automatically when physiological variables breach configurable boundaries — hyperthermia, tachycardia, hypoxaemia, metabolic stress. All state data is exported as self-describing JSON-LD documents anchored to internationally recognised ontology URIs (Uberon, SNOMED CT, VeDDRA), so outputs plug directly into veterinary EHR systems, farm management platforms, and regulatory pharmacovigilance pipelines without translation.

Framework
DigitalSoma framework logo — animal physiology digital twins

DigitalSoma — animal physiology digital twins

DigitalSoma is an open-source Python framework that represents a living animal as a continuously updated digital twin. It implements a three-layer computational architecture — Structural, Dynamic, and Functional — applied to animal physiology, making it the first domain-agnostic digital twin substrate designed explicitly for living organisms across all taxa: livestock, companion animals, wildlife, and aquatic species.

At its core, DigitalSoma treats the animal body not as a static database record but as a continuously evolving computational object. Real-time sensor streams — from wearable collars, implanted biosensors, remote thermal imaging, and laboratory assays — are ingested through a schema-agnostic manifest layer, normalised against a controlled ontological vocabulary, and fed into a composable chain of physiological solvers that infer high-order clinical variables in real time: cardiac output, metabolic rate, thermal comfort index, oxygen consumption, and physiological stress index.

A defining architectural feature is its native alignment with veterinary regulatory standards. All 44 canonical physiological properties are anchored to internationally recognised ontology namespaces — Uberon, SNOMED CT, NCBITaxon, and UCUM — enabling full JSON-LD export. Central to this is the integration of VeDDRA (Veterinary Dictionary for Drug Regulatory Activities), which serves a dual role: in the ontology layer, VeDDRA clinical sign term IDs are registered directly as property descriptors; in the functional output layer, the adverse event screen solver continuously maps the twin’s live physiological state against VeDDRA thresholds, and the veddra_report() method serialises flagged events as a structured report suitable for direct submission to the EMA EVVET3 pharmacovigilance system, the UK VMD, and the FDA Center for Veterinary Medicine.

Architecture

Framework architecture

Click any component to explore its role in the framework.

Wearable
HR · accel · temp
Implanted
bolus · CGM · BP
Remote sensing
thermal · drone · acoustic
Lab assay
blood · urine · eDNA
Manual / custom
field obs · calibration
all sources normalised
Ontology & normalisation layer
vocab.py · canonical_key() · normalise_dict() · to_jsonld()
Uberon · SNOMED · VeDDRA · NCBITaxon · UCUM
canonical keys
DigitalSoma core — soma_api.py
Structural layer
Animal Type Registry
Anatomy map
Normal ranges
build_soma(config)
Dynamic layer
KV store
Time-series log
Threshold events
update_sync(readings)
Functional layer
Built-in solvers
VeDDRA screen
+ custom solver
DAG execution
BYOD manifest — sensor_layer.py · six-field contract: sensor_id · canonical_key · unit · timestamp · value · quality_flag
tool schemas
LLM agentic interface layer
soma_agent.py · SomaDispatcher · 10 OpenAI-compatible tool schemas
Claude · GPT-4 · Gemini
on every update_sync() cycle
State snapshot
all derived vars
Time-series
CSV · JSON · TSL
JSON-LD export
Uberon · SNOMED URIs
VeDDRA report
EMA · VMD · FDA-CVM
LLM response
natural language
Framework

Digital Pedon — soil profile digital twins

The Digital Pedon (DP) is an open-source, zero-dependency Python framework that implements a soil profile as a continuously updated digital twin. Built as the soil-systems counterpart to DigitalSoma, it applies the same three-layer computational architecture — Structural, Dynamic, and Functional — to multi-horizon soil profiles, bridging three persistent gaps in soil science: disconnected models and observations, cross-database interoperability, and the inference gap between raw sensor signals and agronomically meaningful variables.

🤝
Research collaboration
Dr. Ir. Nasem Badreldin — Agroecosystems Laboratory, University of Manitoba
Digital Pedon is developed in collaboration with Dr. Nasem Badreldin (Department of Soil Science, University of Manitoba), whose expertise in digital agroecosystems, precision agriculture, and soil remote sensing has shaped the framework’s scientific foundations.
✉  [email protected]

At its core, the DP treats the pedon not as a static database record but as a continuously evolving computational object. Real-time sensor streams — from in-situ moisture probes, thermocouple arrays, EC sensors, and satellite platforms — are ingested through a schema-agnostic BYOD manifest layer, normalised against 41 canonical soil properties anchored to GLOSIS/FAO and OGC SoilML URIs, and fed into a composable chain of physics-based solvers that infer high-order soil variables in real time: matric potential, unsaturated hydraulic conductivity, vertical flux, thermal diffusivity, and microbial CO&sub2; efflux.

A defining capability is global interoperability. All 41 canonical properties are anchored to internationally recognised ontology namespaces — GLOSIS/FAO, OGC SoilML, and ISO TC 190 — enabling full JSON-LD export. The to_jsonld() method serialises any snapshot as a self-describing linked-data document suitable for direct submission to WoSIS and ESDAC. The LLM Agentic Interface Layer further exposes the framework to Claude, GPT-4, and Gemini, allowing practitioners to interrogate the live soil state in natural language without writing any API code.

Architecture

Framework architecture

Click any component to explore its role in the framework.

SoilGrids v2.0
ISRIC REST API
SSURGO / USDA
SQL · REST · US coverage
IoT sensors
TEROS · CS655 · GS3
Remote sensing
Sentinel-2 · SAR · MODIS
Manual / custom
Lab · calibration fns
all sources normalised
Ontology & normalisation layer
vocab.py · canonical_key() · normalise_dict() · to_jsonld()
GLOSIS/FAO · OGC SoilML · ISO TC 190
canonical keys
Digital Pedon core — pedon_api.py
Structural layer
Soil Type Registry
VG hydraulic params
Threshold config
build_pedon(config)
Dynamic layer
KV store
Time-series log
Threshold events
update_sync(readings)
Functional layer
Built-in solvers
PTF initialisation
+ custom solver
DAG solver chain
BYOD manifest — sensor_layer.py · six-field contract: source_key · canonical_key · unit · horizon_id · depth_cm · sampling_rate_s
tool schemas
LLM agentic interface layer
dp_agent.py · PedonDispatcher · 10 OpenAI-compatible tool schemas
Claude · GPT-4 · Gemini
on every update_sync() cycle
State snapshot
θ · T · EC · ψ · K(θ)
Time-series
CSV · JSON · TSL
JSON-LD export
GLOSIS · WoSIS ready
Threshold alerts
sat · wp · temp · EC
LLM response
natural language
Our Work

Current projects

These are the platforms we are building and deploying right now.

🦋
DigitalSoma
Domain-agnostic animal physiology digital twin framework — any taxa, any sensor, VeDDRA-compliant pharmacovigilance output. Open-source Python, zero dependencies.
●  Live
🌿
Digital Pedon v2.0
A digital twin framework for multi-horizon soil profile monitoring. Open-source, zero dependencies, globally interoperable.
●  Live
🏎
EquiTwin
Personalised bio-digital twin for elite equestrian sport — horse, rider, and racetrack in one connected model.
●  In development
🐄
LiveTwin
Herd-level physiological and welfare monitoring twin for cattle and sheep operations.
●  Coming 2026
🧬
In-Silico Testing Platform
PBPK and organ-level digital twin models for drug and toxin response, aligned with EU/FDA NAM frameworks.
●  Early concept
About

Meet the lead developer

The mind and hands behind Digital Pedon, SoilTalks, and BioTwinRs — bridging soil science, computational biology, and AI.

Dr. Ali Youssef
Lead Developer
Dr. Ir. Ali Youssef
Adj. Prof. “Computational Bio-Ecosystems”
Agroecosystems Lab., University of Manitoba

Researcher and Principal Investigator with over 14 years of experience in Computational Animal & Human Physiology, working across the continuum from cellular-level modelling to population-scale environmental monitoring. Combining advanced sensing technologies (IoT and wearable biosensors) with computational methodologies — including AI/ML and digital twins — to develop interoperable, near-real-time data products for complex biological systems.

A consistent thread throughout his career has been the development of digital and computational alternatives to conventional animal experimentation — including the co-founding of the open Animal Digital Twin Platform (ADTP) and the Digital PhenoLab initiative for non-invasive, continuous monitoring of animal bioresponses. These efforts are grounded in the 3Rs principles and translate directly into in silico experimentation frameworks applicable to veterinary, biomedical, and human-health contexts. Beyond research, a sustained record of industrial collaboration and valorization — leading R&D partnerships with organisations across the sector.

🏛️
Affiliation
Agroecosystems Laboratory, University of Manitoba, Winnipeg, Canada
BioTwinRs Ltd, Leuven, Belgium · DigitalTwinRs Ltd
💻
Get In Touch

Let’s build something together

Whether you are a researcher, investor, veterinarian, or agronomist — if you work with biological systems and need better insight, we want to hear from you.

🏛️
Institution
Agroecosystems Laboratory
University of Manitoba, Winnipeg, Canada
Publications & profiles
We are open to
Investment Research partnerships Technology licensing Custom development Academic collaboration