limix/glimix-core

Fast inference for Generalised Linear Mixed Models.

limix

Genomic analyses require flexible models that can be adapted to the needs of the user.

Limix is a flexible and efficient linear mixed model library with interfaces to Python. It includes methods for

• Single-variant association and interaction testing
• Variance decompostion analysis with linear mixed models
• Association and interaction set tests
• Different utils for statistical analysis, basic i/o, and plotting.

We have an extensive documentation of the library. If you need further help or want to discuss anything related to limix, please, join our forum 💬 and have a chat with us 😃. In case you have found a bug, please, report it creating an issue.

Install

NOTE: We will be maintaining limix 2.0.x for a while, in case you find some missing feature in limix 3.0.x. If that is the case, please, type pip install "limix <3,>=2" in your terminal.

Installation is easy and works on macOS, Linux, and Windows:

pip install limix

If you already have Limix but want to upgrade it to the latest version:

pip install limix --upgrade

Interactive tutorials

• eQTL (requires limix 2.0.x)
• Struct-LMM (requires limix 2.0.x)

Running tests

After installation, you can test it

python -c "import limix; limix.test()"

as long as you have pytest.

Authors

• Christoph Lippert
• Danilo Horta
• Francesco Paolo Casale
• Oliver Stegle

License

This project is licensed under the Apache License License.

glimix-core's Issues

Memory leak due to bound callback

Hi,
LMM passes a bound method as a callback to an optimix Scalar / ndarray-listener. I believe this leads to a memory leak as instantiations of LMM will not be removed by the GC unless _logistic is de-referenced as well.

glimix-core/glimix_core/lmm/_lmm.py

Line 110 in 8cd41ce

logistic.listen(self._delta_update)

A solution could be a wrapper class implementing weak references as described here:
https://www.oreilly.com/library/view/python-cookbook/0596001673/ch05s15.html

Best,
Tobias

ImportError: cannot import name 'FunctionReduce'

Hi

I am trying to use the package, I have installed using conda and get the following error:

from optimix import FunctionReduce ImportError: cannot import name 'FunctionReduce

Although I understand how EP works I am still struggling in the implementation. Would you have a simple example using Logistic or Poisson regression with Gaussian priors please?

_fast_scan_chunk: handle the case where ncovariates==1 and/or offset only

speed up for marc

Memory leak due to the use of functools lru_cache with class methods

Hi,
class method _logdetXX of LMM is cached using functools lru_cache. Calling .fit() or .lml(), which both rely on _logdetXX will prevent the LMM instance to be accessible to garbage collection until the cache is cleared (and listeners are removed, see issue #14).

glimix-core/glimix_core/lmm/_lmm.py

Line 457 in 8cd41ce

@cache

See also
https://stackoverflow.com/questions/33672412/python-functools-lru-cache-with-class-methods-release-object

Best,
Tobias

EP/GLMM lml tailored for linear kernel

High number of trials for binomial might fail to converge

Emmit a warning and clamp down

EP/GLMM posterior updates tailored for linear kernel

Loop over chunks for fast scan

EP/GLMM lml gradient tailored for linear kernel

Low-rank safe GGP

Define ConvergenceError exception to be raised whenever our methods doesn't converge.

Run the same depression experiments with binary QEP