Generic stochastic model #1417

phumtutum · 2021-11-20T22:46:26Z

In this pr:

A generic stochastic model is added (MarkovJumpModel) which uses Gillespie's algorithm
An example of a model that illustrates how simply other stochastic models can be added from now on (Michaelis Menten model)
Added a degradation model for testing as well
Removed old degradation model

Codecov Report

Merging #1417 (27453ef) into master (edbf7c1) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master     #1417   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           87        90    +3     
  Lines         8899      8926   +27     
=========================================
+ Hits          8899      8926   +27

Impacted Files	Coverage Δ
pints/toy/__init__.py	`100.00% <ø> (ø)`
pints/toy/stochastic/__init__.py	`100.00% <100.00%> (ø)`
pints/toy/stochastic/_degradation_model.py	`100.00% <100.00%> (ø)`
pints/toy/stochastic/_markov_jump_model.py	`100.00% <100.00%> (ø)`
pints/toy/stochastic/_michaelis_menten_model.py	`100.00% <100.00%> (ø)`

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update edbf7c1...27453ef. Read the comment docs.

phumtutum · 2021-11-24T20:39:02Z

Now, the old degradation model is removed and the pr is ready for review

MichaelClerx · 2021-12-07T19:11:07Z

docs/source/toy/stochastic/index.rst

+Stochastic Toy Problems
+***********************
+
+The `stochastic` module provides toy :class:`models<pints.toy.stochastic.MarkovJumpModel>`,


I guess initially it's just models?

MichaelClerx · 2021-12-07T19:12:06Z

examples/README.md

@@ -117,6 +117,7 @@ relevant code.
 - [SIR Epidemiology model](./toy/model-sir.ipynb)
 - [Stochastic Degradation model](./toy/model-stochastic-degradation.ipynb)
 - [Stochastic Logistic model](./toy/model-stochastic-logistic-growth.ipynb)
+- [Michaelis Menten model](./toy/model-stochastic-michaelis-menten.ipynb)


Separate category here too?

Yes, I think separate out the stochastic ones into their own subsection of "toy models".

phumtutum · 2021-12-12T21:42:36Z

This pr updates the stochastic degradation model. This model is used in a ipynb in #1413 and therefore changes need to be made to that ipynb in this pr once that pr is merged into master. I will make the requested changes at once with those changes.

pints/toy/stochastic/_markov_jump_model.py

pints/toy/stochastic/_michaelis_menten_model.py

MichaelClerx

Fine with me, if @ben18785 likes it!

ben18785

Thanks @phumtutum -- looks really good. There are a few things that I think would be good to change, then I'm happy to merge.

In the degradation notebook:

I think this line needs rewriting: "averaging the concentration values at each time step, produces a reproducible result" -- I don't think that "reproducible result" is the correct terminology here. I guess you just want to say that the mean of the distribution is that exponential thingy?
"The deterministic mean (from above) is plotted with the deterministic standard deviation" -- I think this is a bit ambiguous. I think you want to say, "We now plot the analytic mean and standard deviation of this process."

In the michaelis-menten notebook:

"The Michaelis Menten model is a stochastic model, more details can be found here.": the MM model actually isn't necessarily a stochastic model (we have a deterministic version of a similar but not same version of the model in the package), and the stuff you link to is about the deterministic model. So, can we say, "This is a stochastic version of the Michaelis-Menten model, and more details about the deterministic analogue can be found here."
Can you also provide a little bit of a description about what this model represents in the first part of this notebook?
"This function, given a set of parameters and times, computes the appropriate times and values using Gillispie's algorithm and then uses interpolation to find the values at the exact times requested" -- it's not clear what the values are (i.e. they are counts of molecules) and "Gillespie" has a typo. I think you also need to explain what the interpolation does and why it is needed. I've noticed "Gillespie" is misspelled throughout the notebook actually.
"Given the stochastic nature of this model we can use multiple simulations to make sure that the runs are covering the same model with the same parameters. Our simulations are close, suggesting we are obtaining accurate simulations." -- this isn't correct, really. All you are showing is that there is not a large amount of stochasticity in these models for these initial values. Indeed, can we show that when we use much smaller molecule counts as initial conditions that we get much more stochasticity?
It would be good to show that the solutions with high molecular counts correspond with the solutions of the corresponding ODE: you could code this up using scipy's odeint within the notebook

ben18785 · 2022-01-18T16:10:09Z

examples/README.md

@@ -117,6 +117,7 @@ relevant code.
 - [SIR Epidemiology model](./toy/model-sir.ipynb)
 - [Stochastic Degradation model](./toy/model-stochastic-degradation.ipynb)
 - [Stochastic Logistic model](./toy/model-stochastic-logistic-growth.ipynb)
+- [Michaelis Menten model](./toy/model-stochastic-michaelis-menten.ipynb)


Yes, I think separate out the stochastic ones into their own subsection of "toy models".

ben18785 · 2022-01-18T16:41:50Z

pints/toy/stochastic/_degradation_model.py

+        if np.any(times < 0):
+            raise ValueError('Negative times are not allowed.')
+
+        variance = np.exp(-2 * k * times) * (-1 + np.exp(k * times)) * self._x0


I checked -- this is correct

ben18785 · 2022-01-18T16:42:04Z

pints/toy/stochastic/_degradation_model.py

+        if np.any(times < 0):
+            raise ValueError('Negative times are not allowed.')
+
+        mean = self._x0 * np.exp(-k * times)


ben18785 · 2022-01-18T16:43:58Z

pints/toy/stochastic/_markov_jump_model.py

+    r"""
+    A general purpose Markov Jump model used for any systems of reactions
+    that proceed through jumps. We simulate a population of N different species
+    reacting through M different mechanisms


"reacting through M different reaction equations"

ben18785 · 2022-01-18T16:45:09Z

pints/toy/stochastic/_markov_jump_model.py

+    2. Calculate the time :math:`\tau` until the next single reaction as
+
+    .. math::
+        \tau = \frac{-\ln(r)}{a_0}


We haven't said what a_0 is.

ben18785 · 2022-01-18T16:45:38Z

pints/toy/stochastic/_markov_jump_model.py

+    3. Decide which reaction, i, takes place using r_1 * a_0 and iterating
+
+    through propensities


This needs quite a bit more explanation, especially since this is the key point of the algorithm

…pints-team/pints into i-1067-generic-stochastic-model

phumtutum · 2022-01-19T00:04:32Z

@ben18785 I have addressed your comments. If there is anything else I should change please let me know.

phumtutum added 5 commits November 12, 2021 18:23

Added generic + michaelis menten

4d4ef3c

jupyter notebook added (not fully working yet)

556e560

change to ipynb

1e99263

removed tau-leaping for now, clarified notebook, minor refactoring

74c8b2e

added ipynb to readme

97d737d

phumtutum added 5 commits November 20, 2021 23:50

added small tests for coverage

1850b2b

Added degradation model(for testing) + tests

b080f09

small fixes

52835a3

small test fix

5098400

moved n_parameters to markovjumpmodel

cfc6c2b

phumtutum requested a review from ben18785 November 21, 2021 23:50

phumtutum added 11 commits November 24, 2021 18:42

partial change(for testing docs)

6d3b874

fix 1

acc6b9d

finished docs

24d1aa8

fix for docs

0a1ffdd

final fix for docs

14ed81a

fix for run-tests.py

4122bcb

more fixes for run-tests.py

f158670

more doc fixes

c4edd3f

testing error source

362567d

attempt to fix

4b12e0e

attempt to fix run-tests

5403208

MichaelClerx mentioned this pull request Nov 30, 2021

ABC Routine #1413

Merged

MichaelClerx reviewed Dec 7, 2021

View reviewed changes

MichaelClerx reviewed Dec 13, 2021

View reviewed changes

pints/toy/stochastic/_markov_jump_model.py Outdated Show resolved Hide resolved

MichaelClerx reviewed Dec 13, 2021

View reviewed changes

pints/toy/stochastic/_michaelis_menten_model.py Outdated Show resolved Hide resolved

Apply suggestions from code review

7f69fda

MichaelClerx approved these changes Dec 13, 2021

View reviewed changes

ben18785 requested changes Jan 18, 2022

View reviewed changes

phumtutum and others added 5 commits January 18, 2022 22:36

addressed most comments, left some documentation

a0d4f24

added explanation for markov jump model

99a56b3

Merge branch 'i-1067-generic-stochastic-model' of https://github.com/…

7c4d14f

…pints-team/pints into i-1067-generic-stochastic-model

license to 2022 + changelog

d14667d

Merge branch 'master' into i-1067-generic-stochastic-model

27453ef

phumtutum requested a review from ben18785 January 19, 2022 00:04

ben18785 approved these changes Jan 19, 2022

View reviewed changes

ben18785 merged commit c7ea751 into master Jan 19, 2022

MichaelClerx deleted the i-1067-generic-stochastic-model branch January 19, 2022 15:38

phumtutum mentioned this pull request Feb 3, 2022

Added additional stochastic models[finished] #1432

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Generic stochastic model #1417

Generic stochastic model #1417

phumtutum commented Nov 20, 2021 •

edited by MichaelClerx

Loading

codecov bot commented Nov 20, 2021 •

edited

Loading

phumtutum commented Nov 24, 2021

MichaelClerx Dec 7, 2021

MichaelClerx Dec 7, 2021

ben18785 Jan 18, 2022

phumtutum commented Dec 12, 2021

MichaelClerx left a comment

ben18785 left a comment

ben18785 Jan 18, 2022

ben18785 Jan 18, 2022

ben18785 Jan 18, 2022

ben18785 Jan 18, 2022

ben18785 Jan 18, 2022

ben18785 Jan 18, 2022

phumtutum commented Jan 19, 2022

		3. Decide which reaction, i, takes place using r_1 * a_0 and iterating

		through propensities

Generic stochastic model #1417

Generic stochastic model #1417

Conversation

phumtutum commented Nov 20, 2021 • edited by MichaelClerx Loading

codecov bot commented Nov 20, 2021 • edited Loading

Codecov Report

phumtutum commented Nov 24, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

phumtutum commented Dec 12, 2021

MichaelClerx left a comment

Choose a reason for hiding this comment

ben18785 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

phumtutum commented Jan 19, 2022

phumtutum commented Nov 20, 2021 •

edited by MichaelClerx

Loading

codecov bot commented Nov 20, 2021 •

edited

Loading