miRW 0.3.0

miRW: A Multi-Omics Random Walk Framework for Sample-Independent Construction of Personalized Protein Interaction Networks in Cancer

miRW

miRW: A Multi-Omics Random Walk Framework for Sample-Independent Construction of Personalized Protein Interaction Networks in Cancer

miRW provides a computational framework for constructing personalized protein interaction networks by integrating multi-omics data and a modulation-adjusted Random Walk with Restart (RWR) algorithm. This tool enables sample-specific network refinement and supports downstream analysis in cancer and other complex diseases.

---

Installation

pip install miRW

Example Usage

Example datasets can be downloaded from:

https://github.com/bioUroZC/miRW/tree/main/miRWsingleLayer/data_example

In cases where no seed gene list is available, users can use single_RW_prepare and single_RW_analysis, which only rely on the expression matrix and the interaction network. High-expression genes (e.g. top 10% per sample) are used as seeds internally.

Running miRW on Example Data (with seed genes)

import pandas as pd
from miRW import single_RWSeed_prepare, single_RWSeed_analysis

# === Step 1: Load example data ===
links = pd.read_csv("Links.csv", index_col=0)
expr = pd.read_csv("exprSet.csv", index_col=0)

# CellMarkers is optional
try:
    seeds = pd.read_csv("CellMarkers.csv")
except FileNotFoundError:
    print("CellMarkers.csv not found; using the first gene as fallback seed.")
    seeds = pd.DataFrame({"symbol": [expr.index[0]]})

# === Step 2: Choose a sample ID ===
sample_id = expr.columns[0]
print("Testing sample:", sample_id)

# === Step 3: Prepare the protein interaction network ===
G, real_w, expr2, seeds2 = single_RWSeed_prepare(seeds, links, expr)

# === Step 4: Run Random Walk with Restart ===
result = single_RWSeed_analysis(sample_id, G, real_w, expr2, seeds2)

# === Step 5: Inspect the output ===
print(result.head())
print("\nmiRW test completed successfully!")

Running miRW without an explicit seed list (expression-driven seeds)

import pandas as pd from miRW import single_RW_prepare, single_RW_analysis

links = pd.read_csv("Links.csv", index_col=0) expr = pd.read_csv("exprSet.csv", index_col=0) sample_id = expr.columns[0]

G, real_w, expr2 = single_RW_prepare(links, expr) result = single_RW_analysis(sample_id, G, real_w, expr2)

print(result.head()) print("\nmiRW test completed successfully!")

Input Data Format

1. Protein Interaction Network (Links.csv)

Required columns:

Column	Description
protein1	First protein/gene in interaction
protein2	Second protein/gene
score	Interaction confidence score

---

2. Expression Matrix (exprSet.csv)

• Rows → Genes
• Columns → Sample IDs
• Values → Expression levels (e.g., TPM, FPKM, normalized counts)

Example:

Gene	Sample1	Sample2	Sample3
TP53	3.12	2.89	3.55
EGFR	1.22	0.98	1.34

---

3. Seed Gene List (CellMarkers.csv, optional)

Required column:

Column	Description
symbol	Seed or marker genes used to initiate RWR

If no seed file is provided, miRW automatically assigns the first gene in your expression matrix as the fallback seed.

---

Output

The function single_RWSeed_analysis returns a DataFrame containing:

Column	Description
Sample	Sample ID
link	Sorted gene pair (e.g., GeneA_GeneB)
miRW-Imp	Importance-adjusted interaction weight
miRW-Flow	Flow-based score derived from RWR

These scores can be used for network-based biomarker discovery, pathway investigation, and integration with downstream multi-omics analyses.

---

Features

• Modulation-adjusted edge weights using expression deviation
• Sparse matrix–optimized Random Walk with Restart
• Personalized network construction for each sample
• Outputs interpretable edge-level importance and flow scores

---

License

This project is licensed under the MIT License.

---

Project Repository

GitHub: https://github.com/bioUroZC/miRW
PyPI: https://pypi.org/project/miRW/