Skip to content

Latest commit

 

History

History
191 lines (139 loc) · 6.56 KB

README.md

File metadata and controls

191 lines (139 loc) · 6.56 KB

KaSim logo

KappaTools

Build Status Join the chat at https://gitter.im/Kappa-Dev/KaSim

KaSim is a stochastic simulator for rule-based models written in Kappa. KaSa is a static analyser for Kappa models.

Kappy is a python library to launch and analyse runs and outputs of Kappa models.

Quick startup

If you are new to Kappa, the easiest way to start experimenting with it is using the webapp.

Editor screenshot with contact map Plot screenshot

Editor screenshot with contact map 2 Stories screenshot

It's available directly in your browser, or for more performance as a downloadable electron app, available for MacOS, Windows and Linux.

Kappa tools are also available as Command-Line Interface programs, which you can either build following the instructions below, or find the binaries included with the electron app in subdir resources/bin.

If you would like to use python to interact with the Kappa tools, the kappy lib is where to look. Here's an example of its usage with ipython

In [2]: import kappy

In [3]: model_text = "%agent: A(x)\nA(x[.]), A(x[.]) <-> A(x[1]), A(x[1]) @ 1e-2,1\n%plot: |A(x[.])|\n%init: 100 A()"

In [4]: kappa_client = kappy.KappaStd()

In [5]: kappa_client.add_model_string(model_text)
Out[5]: [...]

In [6]: kappa_client.project_parse()
Out[6]: [...]

In [7]: kappa_client.simulation_start(kappy.SimulationParameter(.1,"[T] > 10"))
Out[7]: {'simulation_artifact_simulation_seed': 297327779}

In [8]: kappa_client.wait_for_simulation_stop()
Out[8]: [...]

In [9]: kappa_client.simulation_plot()
Out[9]:
  [6.7, 48.0],
[...]
  [0.4, 60.0],
  [0.3, 50.0],
  [0.2, 64.0],
  [0.1, 62.0],
  [0.0, 100.0]]}

See the install instructions to start using kappy.

User manual

See documentation page on kappalanguage.org.

Kappy API documentation is online.

The latex sources of the "older" reference manual (and KaSa one) are available in the man/ directory. To compile the manuel, in addition of a decent LaTeX distribution you need gnuplot and graphviz to generate images (make sure that dot is in the PATH of your OS). To generate the pdf of the manual type

make doc

Installation

Core tools

Released versions come with binaries for MacOS, Windows and Debian derivatives (as Ubuntu). Nightly builds of the master branch are built for these platforms by the continuous integration tools.

If you want or need your own build,

  • Install opam (the OCaml package manager) and initialize it (by issuing opam init)
  • In the source directory, install all the dependencies by opam install --deps-only pinned_libs/default if your OS is OSX or linux, or opam install --deps-only pinned_libs/windows if your OS is Windows.
  • dune build

You can be more fine grained if you only need the command-line tools (and therefore could install less dependencies) by doing opam install --deps-only kappa-binaries followed by make all

If nothing worked for you so far. Well, you're pretty much on your own... Kappa tools depend upon the OCaml native compiler version 4.05.0 or above as well as dune, findlib, Lwt (>= 2.6.0), Re, Fmt, Logs and Yojson libraries. Find any way to install them and you'll be only a make all away from getting Kappa binaries...

Kappy

You should be able to pip install kappy.

  • Under MacOS and linux (and if you're not using a python version so cutting edge that we haven't notice its release yet), wheels that contain the core binaries should be available.
  • For other platforms/python versions, you need to get kappa agents by yourself thanks to the opam package manager by opam install kappa-binaries kappa-agents (or use an externaly hosted REST API)
  • In order to develop in kappy and run all its tests, you need to follow the "get your own build section" above as well as install requests (and future).

Usage

KaSim

In order to run a simulation for 100 time units printing observables values every 0.5 time unit, type

bin/KaSim kappa_file_1 ... kappa_file_n -l 100 -p 0.5 -o data_file

This will produce a data file of 200 point containing the trajectory that was produced during the simulation.

Type:

bin/KaSim --help

for a complete list of options.

Kappy

Do:

import kappy
client = kappy.KappaStd()

to get a kappa client that uses a kappa agent installed locally. Add a string argument specifing the path/to/KaSimAgent to use a specific agent.

A minimal example of usage is:

model = "\
%agent: A(x[x.A]) \
%var: n_0 100 \
%var: k_on 1e-2 \
'rule' A(x[.]), A(x[.]) <-> A(x[1]), A(x[1]) @ k_on, 1 \
%plot: |A(x[.])| \
%init: n_0 A()"
client.add_model_string(model)
client.project_parse()
sim_params = kappy.SimulationParameter(pause_condition="[T] > 100",plot_period=1)
client.simulation_start(sim_params)
client.wait_for_simulation_stop()
results = client.simulation_plot()
client.simulation_delete()
# Rerun with some overwritten values for algebraic variables
client.project_parse(k_on=5e-2,n_0=500)
client.simulation_start(sim_params)
client.wait_for_simulation_stop()
results' = client.simulation_plot()
client.shutdown()

Tests

Launch the core/integration tests by make check.

Regenerate the reference files if you've changed something in the outputs by make build-tests

Launch python tests by nosetests (after having followed the "Get your own build" section).