Developing a blood-based classifier for early detection of breast cancer based on tumor educated platelets. We examine two types of classifiers: a particle swarm-optimized support vector machine (PSO-SVM) and an elastic net (EN).
- Train the PSO-SVM and EN on identical data partitions with age/stage matching
- Additional nested cross-validation approaches for EN
- Differential expression analysis to further investigate cancer-associated transcripts and batch-effect-associated transcripts.
- R, esp
glmnetandcaretfor EN training, andtidyversefor data wrangling - For PSO-SVM: ThromboSeq
- Snakemake
- Esther Lips (Corresponding author): [email protected]
- Kat Moore (Postdoc/bioinformatician): [email protected]
- Marte Liefaard (MD/PhD candidate): [email protected]
- Lodewyk Wessels (Group leader): [email protected]
PSO-SVM input code and output directory locations are HARDCODED. Do not change these unless you want to change the source code. Be warned also that repeated runs of the PSO-SVM will overwrite the previous data unless the output directory is renamed.
├── Snakefile <- Snakemake file with rules to run this project.
|
├── renv.lock <- renv lockfile containing package versioning
│
├── README.md <- The top-level README for developers using this project.
│
├── dataset <- Directory for data storage (excluded from git repo)
│
├── bin <- Location (hardcoded) for PSO-SVM code, and adapted TMM normalization.
|
├── figureOutputFolder <- A (hardcoded) figure folder for PSO-SVM
│
├── pso-enhanced-thromboseq1 <- PSO-SVM output
│
├── Rds <- Interim data files associated with notebooks of the same number.
│
└── src <- Source code for use in this project.
│
└── utils <- Utility scripts, such as rmarkdown.R
Note: Not all directories within the project organization hierarchy are part of the github repository. The omitted directories typically contain raw data, older analyses, or figures.