miidl 0.0.1

Python package for identification of biomarkers powered by interpretable Convolutional Neural Networks

MIIDL

MIIDL /ˈmaɪdəl/, the abbreviation of "Markers Identification with Interpretable Deep Learning", is a package for biomarker screening based on interpretable deep learning.

---

Installation

pip install miidl

or

conda install miidl -c bioconda

---

Features

• One-stop profiling
• Multiple strategies for biological data
• More interpretable, not a "black box"

---

Workflow

1) Quality Control

This is the very first procedure to perform filtering according to the non-missing (observation) rate.

2) Normalization

MIIDL offers plenty of normalization methods to transform data and make samples more comparable.

3) Imputation

By default, this step is unactivated, as MIIDL is designed to solve problems including sparseness. But imputation can be useful in some cases. If needed, there are several methods to choose from.

4) Reshape

In order to apply a CNN model, pre-processed data needs to be zero-completed to a certain length.

5) Modeling

A CNN classifier is trained for discrimination. PyTorch is needed.

6) Interpretation

Captum is designed for model interpretability for PyTorch. This step relies heavily on captum.

---

Tutorials

Welcome! 👋 This guide will provide you with a specific example of how to use this tool properly.