MTV-LMM

MTV-LMM (Microbial Temporal Variability Linear Mixed Model) is a scalable and computationally efficient tool to identify autoregressive taxa (i.e., taxa whose abundance can be predicted based on microbial composition at previous time-points), quantify their temporal effects (i.e., 'Time_explainability') and predict their trajectory in future time points. For more details see Modeling the temporal dynamics of the gut microbial community in adults and infants, Shenhav et al. 2019

Support

For support using MTV-LMM, please email: [email protected]

Software Requirements and dependencies

MTV-LMM is written R. In addition to R 3.4.0 (and higher), it has the following dependencies:

• "dplyr", "vegan" (R packages)

• GCTA_1.26 and higher. We recommend using the latest beta version. After downloading GCTA move it into your home directory using

mv ~/Downloads/gcta_[version] ~/. 

Input format

The input to MTV-LMM is composed of two .csv files :

count table - A matrix of temporal samples by taxa, across multiple hosts. The first row contains the sample ids. The first column includes taxa ids. Then every consecutive column contains read counts for each sample. Note that this order must be respected (see example below).

metadata - The first row contains the headers ('sample_id', 'ind_id', 'Id_num', 'ind_time', 'Sampling_day'). The first column contains the sample ids. The second column contains the subject ids; the third column is the index of each subject (between 1 - number of subjects in the data). The fourth column is the time index where the scale is the experiment's sampling rate (e.g., days, weeks, months). The fifth column is the sampling day (similar to the fourth column, but on the scale of days). Note that these names must be respected (see examples below).

Output format

The output is a matrix of taxa and their temporal effects. Taxa is considered to be 'autoregressive' if the Time_explainability component is significant (p_value_adjusted <= 0.05). Only qc taxa are presented (i.e., taxa that are present in at least 10% of the temporal samples).

VALUE	DESCRIPTION
Time_explainability	estimate of variance explained by the microbial community composition at previous time points
SD_Time_explainability	standard deviation of the time_explainability
Ind_effect	estimate of variance explained by the host at previous time points
SD_ind_effect	standard deviation of the host effect
logL0	log-likelihood under the null hypothesis: no termpral effect
logL	log-likelihood under the alternative
taxa_index	index of the focal taxa in the count table
p_value_adjusted	FDR adjusted p-value of the log-ratio test (null hypothesis: no termpral effect)

Usage instructions

1. Clone this repository ('MTV-LMM') using the terminal.

git clone https://github.com/cozygene/MTV-LMM.git

2. Move the cloned directory into your home directory and change permissions by executing the following commands

mv MTV-LMM ~/. 
cd ~/MTV-LMM/MTV_LMM
chmod +x run_TE.sh
chmod +x run_Prediction.sh

3. Save your input data (metadata and count table) in directory 'Data_files'.

4. Execute

./run_TE.sh [your_gcta_path] [count_table] [metadata_file] 

where

ARGUMENT	DESCRIPTION
your_gcta_path	The path in which the file 'gcta64' is saved (e.g., "~/gcta_1.91.3beta_mac/bin/gcta64"). In windows and ios 'gcta64' file is bin directory (in the GCTA folder). In Linux 'gcta64' is in the main directory
count_table	The full name of your taxa count table, including file type (e.g., otu_table_example.csv)
metadata_file	The full name of your metadata file, including file type (e.g., metadata_example.csv)

Example

To run MTV-LMM on example data:

1. Follow steps 1 and 2 in the usage instructions above.
2. Execute

./run_TE.sh [your_gcta_path] otu_table_example.csv metadata_example.csv

Input -

metadata (first few rows):

sample_id	ind_id	Id_num	ind_time	Sampling_day
E000823.1.8	E000823	1	1.8	54
E000823.2.6	E000823	1	2.6	78
E000823.4.0	E000823	1	4	120
E000823.5.0	E000823	1	5	150
...	...	...	...	...
E001958.2.0	E001958	2	2	60
E001958.2.9	E001958	2	2.9	87
E001958.4.2	E001958	2	4.2	126

count table (first 4 rows and columns):

	E000823.1.8	E000823.2.6	E000823.4.0	E000823.5.0
taxa_1	1	1	2	0
taxa_2	0	0	0	5
taxa_3	0	0	200	0
taxa_4	4	5	0	0

Output (first 2 rows) -

Time_explainability	SD_Time_explainability	Ind_effect	SD_ind_effect	logL0	logL	taxa_index	p_value_adjusted
0.477828	0.07661	1.00E-06	0.014884	3740.364	3776.3	6	0
0.130703	0.066163	0.045816	0.047176	5234.124	5238.217	10	0.002760124