METABIOSYS

Pipeline to generate input data (Microbial abundance profile)

The complete step-by-step instructions to generate the abundance table (input data) from raw metagenomic reads using a combination of tools are provided in the README file. The link is given below:

Download README file

1. Generate the abundance table using the steps provided in the README file
2. Ensure the abundance table has sample names in rows and species names in columns. This table can then be used as input for CRCpred

Sample microbial abundance file created using the above steps:

The microbial abundance can be uploaded here:CRCpred

1. The input table can be uploaded in comma-separated values (.csv) or tab- separated (.tsv) format
2. Upload the input table on the Prediction page of CRCpred using the “Browse” option
3. A sample abundance table containing 13 CRC samples from PRJNA763023, processed using the pipeline provided in the README file, is available as input data. This table can be uploaded by clicking on "Upload Sample File" option on the Prediction page
4. Click on Submit option and CRCpred will initiate the prediction process

1. CRCpred performs feature extraction from the input abundance data and runs three algorithms including RF, XGBoost, and ANN to predict CRC or Healthy status
2. Wait for a while for the results to appear on screen
3. Results for the first 30 samples will be displayed on the web server, while complete results can be downloaded as a .csv file from the download section

The prediction results of each classifier can be downloaded from the download section below:

The output result table, named prediction_result.csv, will contain the following columns:
• Column 1: Sample ID (The IDs for each sample from your input data)
• Column 2: ANN (Predicted label: Healthy or CRC from the Artificial Neural Network (ANN) model)
• Column 3: XGBoost (Predicted label: Healthy or CRC from the XGBoost model)
• Column 4: RF (Predicted label: Healthy or CRC from the Random Forest model)
• Column 5: Overall Probability (The average probability of the sample being predicted as Healthy across all three models—ANN, XGBoost, RF)
• Column 6: ANN Probability (The confidence score of the prediction results from the ANN model)
• Column 7: XGBoost Probability (The confidence score of the prediction results from the XGBoost model)
• Column 8: RF Probability (The confidence score of the prediction results from the Random Forest model)

Probability values closer to 1 indicate a higher likelihood of being Healthy or CRC as predicted by the corresponding model. If all three models agree on the label, it strengthens the confidence in the prediction.