Skip to content

TNBC breast cancer treatment response model #402

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
BVishal-Geek wants to merge 1 commit into
base: main
Choose a base branch
from
Open

TNBC breast cancer treatment response model #402

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
199 changes: 199 additions & 0 deletions flask_backend/models/BreastCancer_SVM_v1/BCO.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,199 @@
{
"object_id": "https://biocomputeobject.org/HIVE_000006/DRAFT",
"spec_version": "https://w3id.org/ieee/ieee-2791-schema/2791object.json",
"etag": "",
"provenance_domain": {
"name": "BreastCancer_SVM",
"version": "1.0",
"license": "https://spdx.org/licenses/CC-BY-4.0.html",
"created": "2025-12-15T19:01:57",
"modified": "2025-12-15T19:42:32",
"derived_from": "",
"contributors": [
{
"name": "Vishal Bakshi",
"affiliation": "George Washington University",
"email": "vishal.bakshi@gwu.edu",
"contribution": [
"createdBy"
],
"orcid": ""
},
{
"name": "Patrick McNeely",
"affiliation": "George Washington University",
"email": "pmcneely@email.gwu.edu",
"contribution": [
"createdBy"
],
"orcid": ""
},
{
"name": "Lori Krammer",
"affiliation": "George Washington University",
"email": "lorikrammer@email.gwu.edu",
"contribution": [
"createdBy"
],
"orcid": ""
}
],
"review": []
},
"usability_domain": [
"This pipeline is intended to support biomedical researchers, data scientists, and clinical informaticians interested in: Extracting the response percentage of breast cancer patients to chemotherapy and combo therapy using a pre-trained SVM model."
],
"description_domain": {
"keywords": [
"python",
"machine learning",
"single-cell RNA-seq"
],
"xref": [],
"platform": [
"VS Code"
],
"pipeline_steps": [
{
"step_number": 1,
"name": "model inference",
"description": "Run inference on the new participant's single-cell RNA-seq data using a pre-trained SVM model.",
"version": "1.0",
"input_list": [
{
"filename": "chemo_model.pkl",
"access_time": "2025-12-15T04:00:00.000Z",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/chemo_model.pkl",
"sha1_checksum": ""
},
{
"filename": "combo_model.pkl",
"access_time": "2025-12-15T04:00:00.000Z",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/combo_model.pkl",
"sha1_checksum": ""
},
{
"filename": "test_script.py",
"access_time": "2025-12-15T04:00:00.000Z",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/test_script.py",
"sha1_checksum": ""
},
{
"filename": "testset_P020.csv",
"access_time": "2025-12-15T04:00:00.000Z",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/testset_P020.csv",
"sha1_checksum": ""
}
],
"output_file": [
{
"filename": "NA",
"access_time": "2025-12-15T04:00:00.000Z",
"uri": "NA",
"sha1_checksum": ""
}
]
}
]
},
"parametric_domain": [],
"io_domain": {
"input_subdomain": [
{
"uri": {
"filename": "chemo_model.pkl",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/chemo_model.pkl",
"access_time": "2025-12-15T04:00:00.000Z"
}
},
{
"uri": {
"filename": "combo_model.pkl",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/combo_model.pkl",
"access_time": "2025-12-15T04:00:00.000Z"
}
},
{
"uri": {
"filename": "test_script.py",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/test_script.py",
"access_time": "2025-12-15T04:00:00.000Z"
}
},
{
"uri": {
"filename": "testset_P020.csv",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/testset_P020.csv",
"access_time": "2025-12-15T04:00:00.000Z"
}
},
{
"uri": {
"filename": "testset_P025.csv",
"uri": "https://github.com/GW-HIVE/PredictMod/blob/main/flask_backend/models/BreastCancer_SVM_v1/testset_P025.csv",
"access_time": "2025-12-15T04:00:00.000Z"
}
}
],
"output_subdomain": [
{
"mediatype": "string",
"uri": {
"filename": "NA",
"uri": "NA",
"access_time": "2025-12-15T04:00:00.000Z"
}
}
]
},
"execution_domain": {
"script": [],
"script_driver": "",
"software_prerequisites": [
{
"name": "Python",
"version": "3.11.13",
"uri": {
"filename": "",
"uri": "https://www.python.org/",
"access_time": "2025-12-15T04:00:00.000Z",
"sha1_checksum": ""
}
},
{
"name": "pandas",
"version": "2.2.2",
"uri": {
"filename": "",
"uri": "https://pandas.pydata.org/",
"access_time": "2025-12-15T04:00:00.000Z",
"sha1_checksum": ""
}
},
{
"name": "numpy",
"version": "2.0.2",
"uri": {
"filename": "",
"uri": "https://numpy.org/",
"access_time": "2025-12-15T04:00:00.000Z",
"sha1_checksum": ""
}
},
{
"name": "scikit-learn",
"version": "2.4.2",
"uri": {
"filename": "",
"uri": "https://scikit-learn.org/",
"access_time": "2025-12-15T04:00:00.000Z",
"sha1_checksum": ""
}
}
],
"external_data_endpoints": [],
"environment_variables": {}
},
"extension_domain": [],
"error_domain": {}
}
118 changes: 118 additions & 0 deletions flask_backend/models/BreastCancer_SVM_v1/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,118 @@
# 🧬 Breast Cancer Treatment Response Prediction
### Single-Cell RNA-seq–based Machine Learning Pipeline

---

## 📌 Project Overview

Triple-Negative Breast Cancer (TNBC) patients typically receive either **Chemotherapy** or **Anti–PD-L1 + Chemotherapy**. However, only a subset of patients respond effectively to each treatment.

This project builds **two machine learning models** using **pre-treatment single-cell RNA-seq data** to:

- Predict response likelihood for **Chemotherapy**
- Predict response likelihood for **Anti–PD-L1 + Chemotherapy**
- Aggregate **cell-level predictions** into a **patient-level response estimate**

The system is designed for **real-world inference**, where each patient may contribute **multiple single-cell observations**.

---

## 🎯 Objective

- Use **pre-treatment immune cell features** to predict treatment response
- Build **separate, treatment-specific models**
- Prevent **data leakage** by respecting patient-level grouping
- Provide a **deployable and interpretable inference pipeline**

---

## 📊 Data Source

**Primary Publication**
*Single-cell analyses reveal key immune cell subsets associated with response to PD-L1 blockade in triple-negative breast cancer*
**PMID:** 33589889
**GEO Accession:** GSE169246

**Data Characteristics**
- Single-cell RNA sequencing (scRNA-seq)
- Pre-treatment immune cells
- Multiple cells per patient
- Cell-level annotations linked to patient response

---

## 🧪 Dataset Structure

Each row corresponds to **one immune cell**.

| Column | Description |
|------|-------------|
| Expression | Aggregate gene expression score |
| nUMI | Total RNA molecules per cell |
| nGene | Number of detected genes |
| percent_mito | Mitochondrial gene fraction (cell stress) |
| percent_hsp | Heat shock protein expression |
| percent_ig | Immunoglobulin expression |
| percent_rp | Ribosomal protein expression |
| PDCD1 | PD-1 gene expression |
| Origin | Tissue source |
| Response | Clinical response label |
| Patient_code | Unique patient identifier |
| Timeline | Pre-treatment / Post-treatment |

---

## 🧠 Modeling Strategy

### Why two models?

Different treatments were trained on **different tissue distributions**:

- **Chemo model**
- `breast`, `liver`
- **Anti–PD-L1 + Chemo model**
- `chest_wall`, `liver`, `lymph_node`

Training a single model would introduce **distribution shift**, so treatment-specific models are used.

---

## 🤖 Models Used

- **Support Vector Machine (SVM)**
- RBF kernel
- Class-weight balanced
- StandardScaler included in pipeline

**Why SVM?**
- Suitable for small datasets
- Handles non-linear decision boundaries
- Stable under limited samples

---

## 🔍 Prediction Logic

1. User uploads a CSV containing **single-cell observations**
2. Data is filtered to **pre-treatment** rows
3. Tissue origin is **one-hot encoded**
4. Each model:
- Receives only the features it was trained on
- Produces **cell-level predictions**
5. Final response score:

percentage = (Number of responder cells / Total number of cells) * 100

---

## Train the model

`python chemo_model_training.py --input "path/to/your/data.csv"`

`python combo_model_training.py --input "path/to/your/data.csv"`

## Run the model

`python test_script.py --input "path/to/your/testdata.csv" --model "chemo_model.pkl"`

`python test_script.py --input "path/to/your/testdata.csv" --model "combo_model.pkl"`
Binary file added flask_backend/models/BreastCancer_SVM_v1/chemo_model.pkl
View file
Open in desktop
Binary file not shown.
Loading