
---
title: "In Silico Perturbation Testing"
description: "How 33MED is engineering the ImmuNav engine to simulate the reactions of Ankylosing Spondylitis Mesenchymal Stem Cells (AS-MSCs) entirely in silico, bypassing wet lab bottlenecks."
date: 2024-05-24
author: "33MED Engineering"
category: "ImmuNav & Computational Biology"
---
The traditional wet lab is a bottleneck. In the pursuit of halting syndesmophyte growth and spinal fusion in Ankylosing Spondylitis (AS), relying exclusively on in vitro perturbation of Mesenchymal Stem Cells (AS-MSCs) restricts iteration speed, explodes costs, and limits observational scale. To engineer programmable immunity and surgical robotics, we require a computational substrate capable of modeling cellular behavior at the speed of software.
Enter **In Silico Perturbation Testing**.
At 33MED, we are building **ImmuNav**, an autonomous biotech infrastructure designed to simulate complex immunological and biomechanical reactions without touching a pipette. By systematically perturbing digital twins of AS-MSCs, we can map transcriptomic shifts, predict protein-ligand binding thermodynamics, and model tissue-scale calcification in real-time.
To achieve this, we have stripped out the bloat of traditional academic bioinformaticsβeliminating cumbersome Python frameworks, fragile message brokers like RabbitMQ, and complex database layers. In their place, we have engineered a highly performant, commercial-safe Node.js, MongoDB, and React ecosystem.
Here is the technical blueprint of the ImmuNav Phase 1 engine.
---
## 𧬠The Core Bioinformatics Tool Matrix
Commercial viability requires navigating a minefield of restrictive software licenses. The ImmuNav engine bypasses restricted tools (like AlphaFold 3's non-commercial wrapper) and orchestrates a matrix of fully open-source alternatives. Our Node.js backend interfaces with these powerful engines using native `child_process` CLI execution, ensuring zero licensing pitfalls and maximum computational throughput.
| Computational Engine | AS-MSC Focus Area | ImmuNav Integration Strategy |
| :--- | :--- | :--- |
| **Chai-1 / Boltz-1** | Multimodal molecular & structure prediction for novel AS-MSC surface receptors. | Executed via standalone pre-compiled CLI binaries, dynamically passing JSON/SMILES inputs. *(Apache 2.0 / MIT)* |
| **GROMACS** | Molecular Dynamics & binding thermodynamics of small-molecule inhibitors targeting IL-17 pathways. | Spawned via native `gmx` commands through Node's `child_process.spawn()` to track real-time thread progress. *(GNU LGPL)* |
| **Bioconductor** | Genomic and transcriptomic pathway mapping of perturbed AS-MSCs. | Executed via standalone `Rscript` CLI to parse RNA/expression datasets. *(Open Source)* |
| **Chaste** | Tissue-scale modeling of pathological osteogenesis (calcification) in the spine. | C++ binaries fired with agent-generated XML payloads mapping cellular parameters. *(BSD License)* |
| **OpenSim CLI** | Simulating mechanical stress on AS-MSC environments in musculoskeletal joints. | Programmatic execution of `opensim-cmd` to validate physical mechanical loads. *(Apache 2.0)* |
---
## ποΈ The Native Node/Mongo Swarm Engine
We do not use dedicated message brokers. To orchestrate our AI agents, MongoDB acts as a durable, atomic task queue using standard queries. This provides a lightweight, highly resilient architecture for our swarm to operate autonomously.
### 1. Atomic Schema Design
Our database architecture tracks agent parameters and manages the task pipeline. We deploy discrete agents for data ingestion, bioinformatics operation, and molecular synthesis.
```javascript
// Collection: agents
{
_id: ObjectId("66654a..."),
role: "BioinformaticsOperator",
status: "active",
globalObjective: "Identify molecular targets to halt syndesmophyte growth and spine fusion.",
humanGuidanceOverride: "Focus explicitly on small-molecule inhibitors targeting IL-17 pathways.",
lastPulse: ISODate("2026-06-08T22:00:00Z")
}
// Collection: tasks (The Lightweight Queue)
{
_id: ObjectId("66654b..."),
agentId: ObjectId("66654a..."),
tool: "gromacs",
status: "queued",
payload: {
inputFiles: ["target_sequence.fasta"],
md_params: { nsteps: 50000, dt: 0.002, integrator: "md" } // Validated by Zod
},
humanSignOff: {
isApproved: true,
actionedAt: ISODate("2026-06-08T22:05:00Z")
},
logs: [],
outputMetadata: { pdbId: "1AKI", finalEnergy: -420000 }
}
2. The Background Processing Loop
In our Simulation Worker, a recursive polling loop queries MongoDB for actionable jobs. It respects a global Pause state and checks for mandatory human approvals before executing computationally expensive physics simulations on the hardware.
async function processQueue() {
// 1. Structural Check: Ensure the swarm isn't globally paused
const globalStatus = await db.collection('settings').findOne({ key: 'swarm_status' });
if (globalStatus?.isPaused) {
setTimeout(processQueue, 5000);
return;
}
// 2. Fetch the next executable task (Must be approved and queued)
const task = await db.collection('tasks').findOneAndUpdate(
{ status: 'queued', 'humanSignOff.isApproved': true },
{ $set: { status: 'processing', startedAt: new Date() } },
{ sort: { createdAt: 1 }, returnDocument: 'after' }
);
if (!task) {
setTimeout(processQueue, 2000);
return;
}
try {
// 3. Native binary execution via Node Child Processes
const binaryProcess = child_process.spawn(getToolBinary(task.tool), task.payload.cliArgs);
binaryProcess.stderr.on('data', (data) => {
// Stream runtime logs directly to the MongoDB task document
db.collection('tasks').updateOne({ _id: task._id }, { $push: { logs: data.toString() } });
});
binaryProcess.on('close', async (code) => {
if (code === 0) {
await handleTaskSuccess(task);
} else {
await db.collection('tasks').updateOne({ _id: task._id }, { $set: { status: 'failed' } });
}
processQueue(); // Instantly jump to the next job
});
} catch (error) {
await db.collection('tasks').updateOne({ _id: task._id }, { $set: { status: 'failed', error: error.message } });
processQueue();
}
}
π‘ Automated Data Ingestion & Computational Immunology
To feed our simulations with the latest AS-MSC literature, ImmuNav features a Data Collection Engine running continuous background tasks.
- PubMed Sourcing: Utilizing NCBI Entrez E-utilities (with a programmatic API key for 10 requests/second), Node fetches literature targeting specific queries (e.g.,
Ankylosing Spondylitis AND syndesmophyte). - arXiv Sourcing: We query the native arXiv Atom/XML feed endpoint (
cat:q-bio.* AND all:ankylosing) to capture cutting-edge preprints. - Gemini Pipeline Integration: Raw XML text is routed directly to the Google Gemini SDK (
@google/generative-ai), instructing the LLM to extract structured biological insights:
const systemPrompt = `You are an expert computational immunologist. Analyze the provided scientific text.
Extract any explicit molecular targets, binding affinities, genetic expressions, or pathways related to Ankylosing Spondylitis.
Return the output strictly in the following JSON format:
{
targetProtein: "string",
associatedCytokines: ["string"],
observedEffect: "string",
confidenceScore: 0.00
}`;
const result = await geminiModel.generateContent([systemPrompt, rawPaperText]);
const structuredData = JSON.parse(result.response.text());
await db.collection('knowledge_base').insertOne(structuredData);
ποΈ Human-In-The-Loop (HITL) Executive Control
Autonomous systems in biotech require absolute executive oversight. To ensure deterministic safety without requiring manual code adjustments, we built the Swarm Management Dashboard (React).
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β IMMUNAV AUTOMATION HUB [ SYSTEM STATUS: RUNNING ] β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β [ π GLOBAL SWARM PAUSE ] [ π RESUME SYSTEM ] [ βοΈ EDIT GLOBAL OBJECTIVE ] β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β ACTIVE AGENTS β
β π€ Agent_Ingestion_1 β Status: [ WORKING ] β Task: Scraping arXiv q-bio β
β π€ Agent_Operator_3 β Status: [ BLOCKED ] β Task: Awaiting Config Approvalβ
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β π¨ AGENT CONFIGURATION APPROVAL GATEWAY β
β Task ID: #98322 β Tool Target: GROMACS Molecular Dynamics β
β Target Interaction: Small-molecule inhibitor binding to IL-17 Receptor β
β β
β [ View Agent-Suggested md_params Payload ] β
β { β
β "integrator": "md", β
β "nsteps": 50000, β
β "dt": 0.002 β
β } β
β β
β βοΈ [ Manually Edit JSON Parameters Inline ] β
β ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β [ β REJECT CONFIGURATION ] [ β οΈ APPROVE & RUN JOB ] β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
When an agent proposes a high-compute simulation run, the task is saved as pending_review. The HITL UI allows 33MED engineers to manually tweak the Zod-validated parameters. Only upon clicking Approve does the system update to queued, passing the job to the local hardware execution loop. Furthermore, a Global Interrupter State ensures the entire swarm can be instantly and safely halted by flipping a boolean in MongoDB, without corrupting ongoing calculations.
π§ͺ Micro-Dataset Testing Protocol: The Local Safety Loop
Developing computational biology infrastructure at this scale demands rigorous testing protocols to avoid exhausting API quotas or crashing hardware under the weight of unoptimized molecular physics.
We enforce a strict Micro-Testing Protocol during development:
- Data Sourcing Sandboxing: We hardcode boundaries into retrieval loops (
&retmax=3for PubMed,&max_results=3for arXiv). This verifies scraping logic, regex sanitizers, and LLM extractions on microscopic datasets before opening the floodgates. - Molecular Modeling Sandbox: Before testing automated Chai-1 or GROMACS scripts on complex human spinal proteins, we validate our architecture using Hen Egg White Lysozyme (PDB ID: 1AKI). Simulating this tiny benchmark structure takes minutes on standard hardware, providing an immediate feedback loop to ensure Node successfully parses stdout strings and exit codes.
- Mechanical Modeling Sandbox: Rather than loading full skeletal rigs, we utilize single-joint pendulums or 2D cellular grid templates found in the OpenSim and Chaste source files. This validates that our React dashboard can accurately render configuration panels and post-run thermodynamic charts before we deploy complex AS-MSC calcification models.
The Future of Autonomous Biotech
By shifting from the wet lab to the ImmuNav engine, 33MED is accelerating the timeline for computational immunology and surgical robotics. In silico perturbation testing isn't just a simulationβit is the foundational infrastructure required to rapidly prototype, validate, and deploy the next generation of interventions for Ankylosing Spondylitis.
We are not just reading the code of human biology; we are engineering the tools to rewrite it.
