AMIVA-F 0.0.32

A FLNc mutation pathogenicity predictor algorithm

AMIVA-F-test package

---

The web-based version of AMIVA-F can be found here:

http://amiva.msp.univie.ac.at/

---

AMIVA-F is a machine learning based algorithm, trained to correlate single point mutations with disease in FLNc.

General Information:

AMIVA-F requires additionally JAVA and PYMOL installed. Tkinter is used for the GUI, which is preinstalled on most python distributions. A step by step tutorial on how to install AMIVA-F is given below for different operating systems.
Currently it is tested on Anaconda3(Windows 10) but more OS and outside of virtual environments (like anaconda) will be tested soon aswell. It was tested as well on Linux Mint, but is not fully documented by now.
This will be soon updated.
Furthermore you can also try to run AMIVA-F outside of a virtual environment like Anaconda.
For that purpose, you would need to make sure that you use the same python interpreter
that PYMOL uses(you need to check which python version your downloaded PYMOL uses).
This will also be documented soon. AMIVA-F alone is relatively small in size, but its required dependencies together with the virtual environment of Anaconda and PYMOL might take up to 3 GB on your computer. If you want to remove AMIVA-F, simply type

pip uninstall AMIVA-F

If you want to update AMIVA-F to the newest version type

pip install AMIVA-F --upgrade

---

Anaconda is a distribution of python programming language that helps with package management and deployment.
Its available for Windows, Linux and macOS.
Package versions in Anaconda are managed by an internal package managment system which does not mess with your local computer package depository.
It further simplifies Path dependency problems and is outlined below for a full installation of AMIVA-F tested on Windows in Anaconda3.

Installation for Windows10(Anaconda3):

If you already installed anaconda, you can skip the next step.

Installation of Anaconda

1. Download Anaconda from https://www.anaconda.com/products/individual
2. Click on the downloaded .exe data.
3. Close all other applications in the background and click Next>
4. Accept by clicking I Agree
5. Select Just Me (recommended) and then Next>
6. Accept the default destination folder and click Next>
7. Select Register Anaconda3 as my default Python 3.8 and click Install

Workflow inside Anaconda3

Setup of the virtual environment

1. Open the anaconda prompt (anaconda3) which you find by entering anaconda prompt into the windows search bar (Found at the left bottom of your screen).

2. This should open a black command line interface where you now need to enter the following command

    conda create -n amivaenv python=3.8
   

3. This creates a new virtual environment with python 3.8 named amivaenv which will be used to install AMIVA and its dependencies without polluting your local pythonspace.

4. You will get a message telling you what is going to be installed. Enter y and press enter.

5. After creation enter:

    conda activate amivaenv
   

6. Which will then activate the new environment.

Getting Java required dependencies

1. Download javabridge from https://www.lfd.uci.edu/~gohlke/pythonlibs/#javabridge
   The site will offer you different versions of javabridge but you need the correct version corresponding to your systems bitness.
   You can check the bitness of your PC by pressing windows key + i together, then navigate to System and then chose About.
   Under Device specifications you will find informations about your system type
   e.g 64 bit operating system, x64-based processor
   If you work on a 64-bit operating system you simply need to download
   javabridge‑1.0.19‑cp38‑cp38‑win_amd64.whl.
   If you work on a 32-bit operating system you simply need to download
   javabridge‑1.0.19‑cp38‑cp38‑win32.whl.
   
   Note: The cp38 part specified the cpython version we use
   (3.8 as we created a virtual environment with python 3.8 before, In case that you want to use another python version you need to select cp37 version for 3.7 and so on).
   
2. Open https://adoptopenjdk.net/
   Under Choose a Version select OpenJDK11 (LTS) and for Choose a JVM select HotSpot
   Click the blue button Latest Release and wait for the download to be completed.
   This will download a .msi file which you should open by left clicking on it (Left bottom of the screen, alternatively found in the Download directory).
   Click Next
   Check the I accept the terms in the License Agreement and click Next
   Attention! Click now the Set JAVA_HOME variable and select Will be installed on local hard drive This is the 3rd row in the directory structure.
   Click Next
   Click Install If everything proceeded normally you managed to install AdoptOpenJDK. Click Finish
   
3. Open http://downloads.gvsig.org/download/gvsig-desktop/runtimes/java/
   Select the appropriate version corresponding to your bitness (if you found before that your bitness is 64-bit you would select For Windows 64 bits (jre-8u72)
   After Downloading the file, leftclick on the .exe file.
   Without altering anything, click Install
   Finish by clicking Close
   This procedure might be tedious but is essential in order to make sure
   WEKA(the machine learning platform which is written in JAVA) can be utitlized from within our python script)
   Congratulations on installing all required JAVA dependencies.
   

Installing everything required for AMIVA-F

If you followed the previous steps, you should now have everything required to make a full installation of AMIVA-F.
Open the Anaconda3 prompt (enter anaconda3 prompt into the windows search bar, found in the lower left corner of the screen).
This will open again the black cmd line.
If you closed this window during the installation process before, you might need to activate again the virtual environment.
Take a look at the beginning of the command line.
If you find (base) at the beginning you need to reactivate your virtual environment.

conda activate amivaenv

Now you should either find (amivaenv) or it was already there because you did not close the terminal before.

1. Open the File explorer and navigate to the Download directory (default directory for downloads).
   Search for the downloaded javabridge-1.0.19 file
   Right click on the javabridge-1.0.19 file and select Properties
   In order to access this downloaded file, we need to specify the path where the file can be found.
   For that reason we could open an editor and simply assemble the path together there which we will then copy/paste afterwards into anaconda.
   You could now open an editor of your choice and do the following:
   Copy the path from the Location Point e.g in my case C:\Users\adm2\Downloads.
   In you case this could look similar like C:\Users\yourname\Downloads.
   Copy this into the editor and add \ and _javabridge-1.0.19-cp38-cp38-win_amd64.whl__(in my case it was this version) which can be found at the top of the window.
   In your editor, the line should now finally look like this:
   C:\Users\adm2\Downloads\javabridge-1.0.19-cp38-cp38-win_amd64.whl
   (again in your case the adm2 name is different)
   
   

2. In the anaconda prompt you now need to enter the following:
   you need to enter pip install and then copy the path you assembled before in the editor

    pip install C:\Users\adm2\Downloads\javabridge-1.0.19-cp38-cp38-win_amd64.whl 
   

If everything worked you will get a message telling you:
Successfully installed javabridge-1.0.19 numpy-1.20.3

3. Now we install AMIVA-F:

pip install AMIVA-F

If everything worked you will get a message telling you:
Successfully installed AMIVA-F-0.0.6 biopython-1.79 freesasa-2.1.0 python-weka-wrapper3-0.2.3

4. We require Pymol:

conda install -c schrodinger pymol

This will ask your permission to install a bunch of files which you accept by entering y and pressing enter.

5. Open AMIVA-F

Open the File explorer and search for your location where you installed Anaconda3.
Usually you will find this under your username
e.g in my case this was adm2 , then select anaconda3, envs, amivaenv,Lib, site-packages, AMIVA-F.
If everything worked well you should see a directory containing the AMIVA-F.py file.
Right click on this file, select Properties and copy the location
e.g C:\Users\adm2\anaconda3\envs\amivaenv\Lib\site-packages\AMIVA-F
Now enter in the command line in anaconda:
this will look slightly different in your case you need to change adm2 to your user name

cd C:\Users\adm2\anaconda3\envs\amivaenv\Lib\site-packages\AMIVA-F	 	  	

and then enter:

python AMIVA-F.py			

This should open now a GUI and you can use AMIVA-F.
For instructions on how to use AMIVA-F see at the bottom of this page.

Linux installation of AMIVA-F (Ubuntu tested, preliminary documentation):

1. Open the linux bash (cmd).

2. Install pip (optional you might have it already)

    sudo apt-get install python3-pip
   

3. Install Jdk (Java development kit)

    sudo apt-get install default-jdk
   

4. Install tkinter (GUI)

    sudo apt-get install python3-tk
   

5. Install PyMol

    sudo apt-get install pymol
   

6. Install AMIVA-F

    pip install AMIVA-F
   

7. Opening AMIVA-F (for me this was the location where AMIVA-F was installed , default)
   In your case you need to switch /user/ and /pythonx.xx/ to your respective name and python interpreter version.

    cd /home/user/.local/lib/pythonx.xx/site-packages/AMIVA-F/
   

Afterwards open AMIVA-F via

	python3.8 AMIVA-F.py

I used python3.8

AMIVA-F Tutorial

Hello, thanks for using the AMIVA-F package in order to predict the pathogenicity for your mutation of interest. AMIVA-F works fully automated and is easy to use, even in the absence of knowledge about the underlying parameters which are used as input for the neural network. In order to make a prediction on your mutation of interest, follow the steps below.

1. AMIVA-F works at the protein annotation level, therefore if you have mutations of interest in the c notation (DNA), look up the corresponding p.notation.
   Once you have your mutation of interest in protein notation, enter it in the entry field location directly above the blue button AMIVA-F Analysis.
   The required input should look like this:

    M82K             
   

This input would correspond to the single point mutation at position 82 in FLNc, where the wildtype amino acid (M, Methionine) is substituted by the mutated amino acid (K, Lysine).
If you by any chance submit a wrong amino acid (the amino acid you specified for the wildtype position is in fact not what you submitted, e.g FLNc position 82 corresponds to methionine, but you wrote S82K, which would correspond to serine), AMIVA-F automatically corrects you and offers you to proceed calculations with the correct amino acid in in place.

2. Click on the button AMIVA-F Analysis.
   This will take a couple of seconds and give you afterwards your prediction.
   After you received your prediction, you can save the prediction and its associated parameters
   used as input for AMIVA-F in simple text format by clicking "Save Mutation prediction".
   This will prompt you in order to choose a directory where the outputfile will be saved.
   In this outputfile you will not only find the prediction together with its input parameters,
   but furthermore also details about the associated training set accuracy and many more statistically interesting things.
   

3. (Optional) If you want to make another mutation, simply delete (or click clear) and enter your new mutation into the entry field at the beginning.
   It is sufficient to simply delete the old mutation and enter the new mutation and then to click AMIVA-F Analysis.
   The old parameters will be deleted and everything will proceed as if you just started AMIVA-F.
   When you are done, simply leave AMIVA-F by clicking Exit AMIVA-F which will turn off the Java Virtual Machine in the background.

Additional Information: For additional information about the posttranslational sites/ binding partners which are used to feed AMIVA-F, there are 2 buttons where you can take a look into them.

Regarding performance prediction accuracy:
Currently this seems to be a bit buggy (Given that it only gives you only the summary of 1 out of the 10 crossfold validated runs (Instead the average of 10) and can take values a bit lower or higher than reported in the paper).
If you take the trainingset to WEKA (a desktop installed version) and do the cross validation there, you will find the exact numbers specified in the publication of 78.6% accuracy.
For your prediction it can happen that you will find prediction values of 76 - 80%,
but these are not true values but instead single models from the 10x cross validation procedure.
We are working to fix this issue in order to provide the average accuracy of 78.6 %, which results after averaging all 10 models generated in cross validation.

More information can be found at the full tutorial inside the package.