Examples

aggmap usage examples

MNIST reconstruction from pixel random permutation

• Introduction
• Step0: Import AggMap and Orignal MNIST data
• Step1: MNIST pixel random permutation
• Step2: AggMap pre-fitting on training set
• Step3: AggMap transformation on training and test test
• Step4: AggMap visualization

Fashion-MNIST reconstruction from pixel random permutation

• Introduction
• Step1: MNIST pixel random permutation
• Step2: AggMap pre-fitting on training set
• Step3: AggMap transformation on training and test test

Pick up stage-specific genes in cell cycle data

• Introduction
• Read data and pre-fit on AggMap
• Transform the data by AggMap
• Multi-channel Fmaps
• Single-channel Fmaps
• Visulization
  • 01.scatter plot
  • 02.grid plot
  • 03.tree plot

Metagenomic deep learning and biomarker discovery

• 1. Introduction
  • 1.1 MEGMA introduction
  • 1.2 Metagenomic cross nation datasets and tasks
  • 1.3 MEGMA fitting and AggMapNet training strategy
• 2. MEGMA Training & Transformation
  • 2.1 Fitting MEGMA on metagenomic abundance data of all countries
    • 2.1.1 Read and preprocess data for MEGMA
    • 2.1.2 MEGMA initialization, fitting and dump
    • 2.1.3 MEGMA loading and 2D-microbiomeprints transformation
    • 2.1.4 MEGMA 2D-microbiomeprints visulization
    • 2.1.5 Well-trained MEGMA to transform the abundance data of each country
  • 2.2 Fitting MEGMA on metagenomic abundance data of one country only
    • 2.2.1 Read and preprocess data for MEGMA
    • 2.2.2 MEGMA initialization & fitting
    • 2.2.3 MEGMA 2D-microbiomeprints transformation
    • 2.2.4 MEGMA Fmaps visulization
    • 2.2.5 Transform the abandance data of the rest countries by country-specific MEGMA
    • 2.2.6 Fitting country-specific megma for all countries
  • 2.3 Discussions & conclusions on MEGMA 2D-microbiomeprints
• 3. Training the CRC detection models based on MEGMA Fmaps
  • 3.1 Training and test AggMapNet on overall MEGMA Fmaps
  • 3.2 Training and test AggMapNet on country specific MEGMA Fmaps
  • 3.3 Comparing the STST performance and discussion
• 4. Important microbial marker identification
  • 4.1 Calculate the global feature importance
    • 4.1.1 GFI for model trained on overall MEGMA Fmaps
    • 4.1.2 GFI for model trained on country specific MEGMA Fmaps
  • 4.2 Generate the explaination saliency map
    • 4.2.1 Saliency map for overall MEGMA Fmaps
    • 4.2.2 Saliency map country specific MEGMA Fmaps
  • 4.3 Global feature importance correlation
  • 4.4 Discussions and conclusions on saliency map
• 5. Toplogical analyisis on the important microbes
  • 5.1 Plotting the the embedded and arranged microbes
  • 5.2 Plotting the linear assignment of the embedded microbes
  • 5.3 Fetching the optimized toplogical graph and clustering
  • 5.4 Exploring the toplogical relationship of the important microbes
• 6. Exploring the embedding & grouping in MEGMA
  • 6.1 Microbial embedding
    • 6.1.1 Manifold embedding
    • 6.1.2 Ramdom embedding
  • 6.2 Microbial grouping
    • 6.2.1 Hierarchical clustering tree based grouping
    • 6.2.2 Taxonomic tree based grouping