This model was built using mlr3verse version 0.2.8 and paradox version 0.11.1. Please ensure that the correct versions of these R packages are installed before using the model, as mismatched versions may lead to unforeseen errors.
A machine learning model for CRC diagnosis
The model has been trained on extensive clinical ELISA data to achieve high-performance discrimination of colorectal cancer. It shows great potential for early colorectal cancer diagnosis.
In brief, to diagnose new samples, the concentrations of CTTN, HNRNPK, and PSMC6 in plasma EVs must first be measured using the ELISA. The concentration values are reported in ng/mL. These three protein concentrations are then input into the ColonTrack model, which calculates a score based on these values. If the score is greater than or equal to 0.5, the sample is classified as CRC; otherwise, it is classified as non-CRC.
The ColonTrack.rds file is a machine learning model file built using the R mlr3 package. The trails_data.csv file contains the data used for model training.
# R code example
ColonTrack <- readRDS("path/to/ColonTrack.rds")
trails_data <- read.csv("path/to/trails_data.csv")
predict = ColonTrack$predict_newdata(newdata = trais_data)
trais_data |>
dplyr::mutate(ColonTrack = predict$prob[, 1])> trais_data |>
+ dplyr::mutate(ColonTrack = predict$prob[, 1])
Sample CTTN HNRNPK PSMC6 ColonTrack
1 Sample1 0.11 0.00 0.72 0.0093333333
2 Sample2 0.13 0.60 0.94 0.0042666667
3 Sample3 0.11 1.14 0.56 0.0020000000
4 Sample4 0.15 1.65 0.44 0.0056333333
5 Sample5 0.26 0.10 0.60 0.0004000000
6 Sample6 0.12 0.00 1.23 0.0508333333
7 Sample7 2.77 1.86 0.56 0.4271333333
8 Sample8 0.13 0.51 0.54 0.0059333333
9 Sample9 0.14 0.00 0.92 0.0162095238
10 Sample10 0.11 0.62 0.84 0.0000000000
The original ColonTrack model was trained using a dataset from a single center, consisting of 123 CRC patients and 118 non-CRC controls. Despite the model showing an accuracy greater than 90% in over 800 samples, the limited sample size and single-center origin of the dataset may introduce some bias, and the model may not fully represent the broader characteristics of CRC. Therefore, its generalizability and robustness require further investigation.
To improve upon this limitation, we trained a new model, ColonTrack-Pro, using ELISA data from four centers, with 557 CRC and 545 non-CRC samples. ColonTrack-Pro is also available for open-source use on GitHub. While this model has not yet been validated with external cohorts, it holds promise for broader diagnostic applications and serves as a valuable reference for future validation efforts.
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, subject to the following conditions:
This Model is for educational and research purposes only. Commercial use is prohibited without explicit written permission from the author.
(MIT License content follows)