AI Powered Photoswitchable Screen (AIPyS) Version 2

Introduction

AIPyS V2 is an AI-driven platform enhancing the capabilities of photoswitchable genetic CRISPR screen technology. Utilizing advanced algorithms like U-net and cGAN for segmentation and employing Bayesian inference for differential sgRNA abundance analysis, AIPyS offers precise detection of single cells and subcellular phenotypes in microscopy images. It integrates Numpy, scikit-image, and scipy for parametric object detection and leverages the PyMC3 library for statistical modeling. For interactive data exploration and visualization, the platform is deployed online via Plotly-Dash.

For detailed insights, visit the Documentation.

Quick Installation Guide

AIPyS supports Windows environments and necessitates Python 3.8. For seamless operation with machine learning components like PyTorch and Cellpose, please align CUDA and cuDNN versions meticulously.

• Conda Installation: Conveniently install using the provided environment.yml to configure both Python and CUDA/cuDNN dependencies accurately.

  conda env create -f environment.yml
  conda activate aipys_env
  

• PIP Installation: For environments where Conda is unavailable, use pip while ensuring correct CUDA/cuDNN configurations.

  pip install AIPySPro
  

Check installation:

aipys --version

Highlighted Features

Segmentation and Analysis

• Parametric Segmentation: Enhances R-based code for effective segmentation using scikit-image.
• Deep Learning Segmentation: Incorporates U-net and cGAN models for cutting-edge segmentation accuracy.
• Granularity Analysis and Classification: Utilizes logistic regression and CNN classifiers trained on meticulously segmented cell images for precise phenotype classification.

Deployment and Integration

• Nikon-nis Elements Integration: Employs AIPyS for advanced image processing, offering streamlined deployment capabilities for Nikon-nis Elements jobs module.
• Interactive Data Visualization: Leverages Plotly-Dash for an immersive data visualization experience, allowing users to interactively explore analysis outcomes.

Bayesian Model Training for Granularity Analysis

• Utilizes Bayesian inference to train models capable of discerning intricate subcellular phenotypes, contributing significantly to the understanding and characterization of genetic modifications impacting cell morphology.