👨‍🔬DNA forensics

In this tutorial, you will learn how to use the Coretex platform for building a model capable of predicting the body site from which the sample was collected based on DNA analysis of the microbiome composition. This model can then be used for forensics purposes in crime labs.

We will use a curated dataset from Microbiome Atlas Project (MAP).

Data can be downloaded from this link: https://microbeatlas.org/index.html?action=download.

You have to download 2 required files:

• samples.info - includes metadata about all samples

• sample.[level].mapped.qz - includes OTU tables with mapped OTU counts for each sample

Different levels represent similarity cutoffs (90%, 94%, 96%, 97%, 98%, and 99%) with SSU rRNA reference composed of 1.5 million full-length sequences obtained using the MAPseq tool.

Import data to Coretex

The first step is to upload the MAP data you want to train your model on to Coretex.

You will have to prepare the structure of the previously downloaded files for upload. This is what you should have before uploading:

• sample.info

• sample.[level].mapped.qz -unzip this file to get sample.[level].mapped

Select both files and compress them.

Since the archive is big (at least 2GB) the upload will take some time, so please be patient.

To learn more details about various ways Coretex lets you move data around (especially large datasets) please visit Dataset upload page. We will use the Coretex CLI to upload large files.

Let's start with the command:

coretex dataset --help

CLI tool - Coretex dataset commands

We can see all the options that can be performed on the dataset. Our focus in this tutorial is creating datasets and uploading samples.

coretex dataset create --help

CLI tool - Coretex create dataset commands

Arguments and flags that we need to send to upload the dataset successfully are displayed by running the previous command.

Let's create a dataset!

coretex dataset create Atlas 8770 11 "/Users/user/Downloads/Atlas Data.zip"

CLI tool - Creating dataset and uploading samples

To run Microbiome analysis with previously uploaded data follow these steps:

1. Find Microbiome analysis Project and select it,

   

   Preview of first step

2. Go to the Workflows screen and open 'microbiome-forensics-workflow',

   

   Preview of Runs screen
3. Fill the necessary fields and Run parameters as explained in the next picture,

   

   Explanation of Run screen for Microbiome analysis
4. Click on Run Task,

5. On the Runs screen find your Run and click on the right button to see more information about your Run.

   

   Preview of recent runs on Run screen
6. On the Run details screen click on the Artifacts button to see predictions of validation after the run is marked as Done.

   

   Preview of run details screen
7. In the artifacts list, you can find a .csv file that contains model predictions.

   

   Preview of artifacts on run details screen

That would be it for this tutorial! To find out more about Coretex check out our other examples of Demo Runs.

Parameters Information

Name	Value	Data type	Required
dataset	null	dataset
validation	false	bool
trainedModel	null	int
datasetType	0	int
taxonomicLevel	1	int
sampleOrigin	human	list[str]
percentile	100	int
quantize	false	bool
sequencingTechnique	AMPLICON/SHOTGUN/WGS	list[str]
cache	true	bool
learningRate	0.3	float
epochs	100	int
earlyStopping	0	int
validationSplit	0.2	float
useGpu	false	bool
logLevel	4	int