MeDUSA

MeDUSA is modular, customizable, and user friendly toolset in R to ease the data processing of Direct Infusion Untarged Single-cell Analysis.

Running

Via Dockerhub

Included is a dockerfile that compiles all R requirements, some common compound databases, and an Rstudio. docker pull thefollyllama/medusa docker run -e PASSWORD=medusa -p 8787:8787 -v .:/home/rstudio/local thefollyllama/medusa in a browser navigate to "localhost:8787" usr:rstudio pwd:medusa

Via Docker

It is suggested to run via dockerhub, as building the image can take over an hour. However, local build instructions can be found in the Dockerfile

Data Structures

MeDUSA's modularity is achived via common data objects that are interchangeable.

MZ-OBJ: The standard "MZ Object"

DataFrame where:

• Rownames = Float: MZ
• Colnames = String: sample_names
• Data = Float: Intensities

MzT-OBJ: An "MZ Object" rebranded to represent scan-time.

DataFrame where:

• Rownames = Float: MZ
• Colnames = Float: ScanTime
• Data = Float: Intensities

mzLog-OBJ: A log2 of "MZ Object."

DataFrame where:

• Rownames = Float: MZ
• Colnames = String: sample_names
• Data = Float: log2(Intensities)

Programatically: the methods for [mz-obj, mzT-obj, mzLog-obj] are all interchangeable....proceed with caution.

mzLong-OBJ: A pivot longer of an "MZ Object"

DataFrame where columns are:

• Float : MZ
• String: ScanTime
• Float : intensity

METADATA: The standard "MZ Metadata object"

Dataframe required columns are:

• Integer: measurement
• String : sample_name (Must match above)
• String : type (i.e. cell, media_cell, solvent)
• String : phenotype Additional columns could be helpful. Such as:
• Integer: time
• Integer: plate
• Integer: cell_count
• String : sampler
• String : polarity
• String : sampling_day
• String : filtered_out

R Project Structure

MeDUSA's files and methods are named to quickly identify input and output expectations.

Primary files:

file_name : [suggested order][input object type][description].R i.e. "2_mz_quality_check.R"

public function : [input_object]_[category]_magic suggested method to handle everything in the file category (i.e. "mz_post_magic")

public function : [input object][subcategory][description] (i.e. "mzlog_analysis_fold")

private funciton : [type].[description] (i.e. "extract.binning")

Support files

file: z_[description].R

• common methods written by ehetzel file: zz_[description].R
• common methods written by others

Support Project Structure

Also in this repo are Rproj adjacent tools

docs: JOSS paper materials examples : a small set of data and an example of ideal usage scripts : mostly data massaging scripts to provide compound data in the docker container validation: TBD Dockerfile: An optimized docker image ~2.5gb Dockerfile_unoptimized: More debug friendly dockerfile ~12gb