Detecting context dependent mutation rates from NGS data.
We designed a statistical method that allows for detection and evaluation of the effects of different motifs on mutation rates.

"CIPI: A Bayesian framework for inferring the influence of sequence context on point mutations" ,Guy Ling, Danielle Miller, Rasmus Nielsen, Adi Stern

A jupyter notebook for running the code in available here with two optional modes:

• simulate
• real data inference

Input frequency file can be either a csv file or a tab delimited file (.freq):
Note that the simulations result have an additional line :#moran model results

#moran model results
Pos    Base    Freq    Ref    Read_count    Rank    Prob
1    A    0.9826    A    10000    3    1.0
1    C    0.0012    A    10000    0    1.0
1    G    0.0058    A    10000    1    1.0
1    T    0.0104    A    10000    2    1.0
2    A    0.9852    A    10000    3    1.0
2    C    0.0053    A    10000    1    1.0
2    G    0.0030    A    10000    0    1.0
2    T    0.0065    A    10000    2    1.0

Generating frequency files from raw NGS data (fastq files) can be done by our AccuNGS pipeline
"AccuNGS: detecting ultra-rare variants in viruses from clinical samples." (Gelbart, M., et al. 2018).
code available here.

The script creates a folder named resultSummary with two types of files:

• MutationFilename_allChainsSummary.csv - all MCMC chains
• MutationFilename_bestChainsSummary.csv - the final file for analysis

Each file should include all possible motifs and their final beta and gamma:

betaMean	gammaMean	betaSD	gammaSD	topBeta	bottomBeta	topGamma	bottomGamma	name
-0.035037904	0	0.370295511	0	0.610536338	-0.644909518	0	0	P4__A
0.111123401	0	0.253151736	0	0.376907698	-0.847964432	0	0	P4__C
1.882467488	1	3.86E-11	0	1.882467488	1.882467488	1	1	P4__G
0.048830458	0	0.32543011	0	0.713381789	-0.511792261	0	0	P4__T
0.159739611	0	0.191817238	0	0.505363867	-0.320087558	0	0	P1_P4__AG
-0.11833743	0	0.259224332	0	0.455072615	-0.474370422	0	0	P1_P4__AT
-0.076704198	0	0.190523941	0	0.304878287	-0.463398403	0	0	P1_P4__CG

Here we can see that the motif P4__G is significanct as gammaMean = 1.

Also, a gamma and beta posterior plots will be saved. For only presenting figures go to GammaPriorTest.py and under outputAnalysis set plot=True .

Running times might differ as a factor of genome length (up to an hour for 7k bases).
For multiple simulations we recommand running a distributed script and not using the attached jupyter notebook.

Python 2.7 with the following packages installed:

• numpy
• pandas
• Bio
• pymc3
• theano
• matplotlib
• seaborn
• tqdm