RunMat for life sciences

Write metabolic models, signal analyses, dosing models and device checks in code that reads like the equations in your methods section, on a runtime that is free for teaching, academic research and smaller companies.

Talk to us

Write your model the way it reads in your paper

The MATLAB language reads like the equations in your methods section, so anyone in your lab can check the code against the paper, line by line.

RunMat runs it, and checks shapes and units before anything runs.

Linear algebraDifferential equations

max⁡  c⊤v,    Sv=0\max\; c^{\top} v, \;\; S v = 0
v = linprog(-c, [], [], S, b, lb, ub);
x˙2=kax1−kex2\dot{x}_2 = k_a x_1 - k_e x_2
dx = [-ka*x(1); ka*x(1) - ke*x(2)];
E=Emax⁡ CnEC50 n+CnE = \dfrac{E_{\max}\, C^{n}}{EC_{50}^{\,n} + C^{n}}
E = Emax * C.^n ./ (EC50^n + C.^n);
1N−1∑(NNi+1−NNi)2\sqrt{\tfrac{1}{N-1} \textstyle\sum (NN_{i+1} - NN_i)^2}
rmssd = rms(diff(nn));

Hand your methods to every student and collaborator

Most accredited biomedical engineering programs require this language, so your incoming students already read it. The RunMat runtime is free for teaching and academic research: a new student runs the lab's code on day one, and a thesis still runs after graduation.

A collaborator at another institution opens the project in a browser. You both point at the same alpha peak, change the window and see the new spectrum together.

Run one analysis across every subject in a cohort

A script that works on one subject's recording, or one knockout, soon has to run on a whole cohort or every reaction in a genome-scale model. Spreading that over machines usually means someone writes and maintains a cluster job.

In RunMat you change for to parfor. The same script runs across every core of your laptop, on its GPU or on your institution's cluster.

ParallelGPU

% HRV for every subject in a study cohort
files = dir('cohort/*.csv');
sdnn = zeros(1, numel(files));
parfor k = 1:numel(files)
    sdnn(k) = hrv_sdnn(readmatrix(fullfile('cohort', files(k).name)));
end
low = files(sdnn < 50);
LaptopRemote serverLARGE CLUSTER

A deterministic engine designed for AI models to use in controlled loops

RunMat is a deterministic engine: the same script and data give the same result every time. It is designed for AI models to use in a controlled loop, for work like the analyses on this page.

The AI model writes the code in the same notation as your equations. RunMat checks its types, shapes and units and sends any error back for the model to fix, then runs it. The model reports the result and how it was computed, and every number in that answer came from RunMat, so it can go in your paper.

Use Claude Code, Codex, any MCP client or the agent built into RunMat, on the model your institution already uses.

AI integrations

Works withClaude CodeCodexAny MCP client

AI modelSets up and interpretsRunMatChecks and runs the studystudyerrors and results
  1. ModelWrites cantilever.fea
  2. RunMatline 28: traction must be force per area, found N
  3. ModelFixes the load unit: N → Pa
  4. RunMatSolved · tip deflection 0.199 mm, beam theory 0.200 mm
  5. ModelReports the results and how they were computed

Know a wearable's skin temperature at full power before you build the prototype

A device on the skin has to stay under the contact-temperature limit at full power.

In RunMat the thermal model is its two heat balances, written as they read on paper. Change the power or the case and run it again, and each run is kept with the inputs that made it, ready for your design review. When the question needs the real part, a heat or stress study of its CAD file runs in the same project.

Differential equationsSimulation in RunMat

wearable-temperature.m
% A wearable on the wrist: 0.35 W from its electronics, a two-node model of the
% board and the skin-contact case, against a 43 C limit for skin contact.
P = 0.35; Cb = 4; Cc = 12; % W, J/K board, J/K case
Rbc = 25; Rskin = 60; Rair = 90; % K/W board-case, case-skin, case-air
Tskin = 33; Tair = 25;
f = @(t, T) [(P - (T(1) - T(2))/Rbc) / Cb;
((T(1) - T(2))/Rbc - (T(2) - Tskin)/Rskin - (T(2) - Tair)/Rair) / Cc];
[t, T] = ode45(f, linspace(0, 3600, 600), [Tair; Tair]);
Rpar = 1 / (1/Rskin + 1/Rair);
steady = (Tskin/Rskin + Tair/Rair + P) * Rpar;
fprintf('case %.1f C after an hour, %.1f C at steady state\n', T(end, 2), steady);
fprintf('margin to the 43 C skin-contact limit %.1f K\n', 43 - max(T(:, 2)));
Terminal
$ runmat wearable-temperature.m
case 42.4 C after an hour, 42.4 C at steady state
margin to the 43 C skin-contact limit 0.6 K

Keep patient data on your servers

RunMat can run on your own servers, fully offline if you need to, so patient recordings and trial data stay inside your institution.

Self-hosted deployment

Examples

Bar chart of growth kept after each single-reaction knockout, rendered by RunMat

Flux balance and knockouts

Which reactions a cell can lose and still grow.

Beat-to-beat intervals over two minutes of ECG, rendered by RunMat

Heart-rate variability from an ECG

R peaks, beat-to-beat intervals, SDNN and RMSSD.

Plasma concentration over eight doses reaching steady state, rendered by RunMat

Dosing to steady state

A one-compartment oral model, 500 mg every 12 hours.

Assay responses against dose with a fitted Hill curve, rendered by RunMat

Dose-response fit

A four-parameter Hill fit to a triplicate assay.

EEG power spectra with eyes open and closed, rendered by RunMat

EEG alpha power

The 10 Hz rhythm with eyes open and closed.

Board and case temperature of a wearable over an hour against the 43 C limit, rendered by RunMat

Wearable skin temperature

A device on the wrist against the 43 C skin-contact limit.

What RunMat does not do

  • Validation of RunMat itself for FDA submissions or under IEC 62304
  • DICOM viewing, PACS and whole-slide imaging
  • Sequence alignment and genomics pipelines
  • Clinical decision support

Start here

  1. 01

    Open the flux balance example in your browser. No account needed.

  2. 02

    Run three of your lab's scripts and compare their results and run times.

  3. 03

    Talk to us about your lab, your data and where it runs.

Signal processingDifferential equationsCurve fittingStatistics