180-Day VTE Risk Prediction Model (Random Survival Forest)

This repository provides the trained Random Survival Forest (RSF) model used in the manuscript:

"Development of novel plasma proteomic biomarkers for cancer-associated thrombosis in an advanced cancer cohort."

Repository Contents

• final_rsf_vte_model.rds --- Trained RSF model object
• demo_load_model.R --- Example script demonstrating model loading and prediction

---

Software Requirements

Model was trained using:

• R version 4.5.1
• randomForestSRC 3.5.1
• survival 3.8-6

A complete computational environment snapshot is provided in session_info.txt.

To ensure compatibility, we recommend using the same or newer versions.

Install required packages:

install.packages("randomForestSRC")
install.packages("survival")

Load packages:

library(randomForestSRC)
library(survival)

---

Model Description

Parameter	Value
Model Type	Random Survival Forest
Outcome	Time to VTE
Prediction Target	180-day VTE risk
Trees	1000
mtry	74
nodesize	10

---

How to Use

Place final_rsf_vte_model.rds in your working directory and run:

library(randomForestSRC)
library(survival)

model <- readRDS("final_rsf_vte_model.rds")

# Example prediction on training data
pred <- predict(model)

t180 <- 180
idx180 <- which.min(abs(pred$time.interest - t180))
risk_180 <- 1 - pred$survival[, idx180]

head(risk_180)

For new data, ensure that:

• All predictor variables match those used in training

---

Important Notes

• The model requires the same predictor variables and preprocessing used during training.
• Raw training data are not publicly available due to patient privacy restrictions.

---

Reproducibility

The model object contains:

• Full forest structure
• Predictor names
• Hyperparameters
• Survival time grid

Users may inspect:

model$call
model$xvar.names

---