MEchanistic Clustering - Treatment eXposure Framework
mectx implements the MEC-TX framework for encoding, clustering, and survival analysis of real-world oncology treatment timelines. It was developed for registry-based cohorts such as the ORIEN AVATAR dataset.
Install the released version from CRAN:
install.packages("mectx")Or the development version from GitHub:
# install.packages("devtools")
devtools::install_github("spakowiczlab/mectx")| Function | Description |
|---|---|
tx_normalize() |
Normalize raw medication records to a time grid |
tx_intervals() |
Compute treatment intervals per patient |
tx_cluster_surv() |
K-means clustering in PCA space with survival output |
tx_lines() |
Assign line-of-therapy labels |
tx_pooled_analysis() |
Compare survival across treatment groups |
tx_duration() |
Summarise treatment exposure duration by group |
tx_compare_groups() |
Statistical comparison across patient groups |
dominant_exclusive() |
Assign mutually exclusive dominant regimen per patient |
get_focus_cohort() |
Filter cohort by focus treatment type |
tx_focus_dt() |
Build digital-twin timeline for focus treatment |
library(mectx)
# Step 1: Normalize raw medication data
norm <- tx_normalize(raw_medication_df)
# Step 2: Compute treatment intervals
intervals <- tx_intervals(norm)
# Step 3: Cluster patients by treatment pattern
clustered <- tx_cluster_surv(norm, meta_df)
# Step 4: Assign line-of-therapy
lines <- tx_lines(intervals)
# Step 5: Pooled survival analysis
results <- tx_pooled_analysis(intervals, meta_df, group_var = "CAlevel")
# Step 6: Compare treatment duration by group
duration <- tx_duration(intervals, meta_df, group_var = "CAlevel")
The canonical MEC-TX pipeline order:
raw data └─ tx_normalize() └─ tx_cluster_surv() └─ tx_intervals() └─ tx_lines() └─ tx_pooled_analysis() └─ tx_compare_groups() └─ tx_duration()
If you use mectx in your research, please cite: need to double check
Dhrubo and Spakowicz (2026). MEchanistic Clustering - Treatment eXposure Framework for real-world oncology treatment timeline analysis.
MIT + file LICENSE