MitochondriaAI: Sex-Based Analysis of Mitochondrial Function in Pancreatic Beta Cells

Authors

Fahd Qadir (Dragonmasterx87)
FMJ Lab, Tulane University School of Medicine

Charles Dana (AlgorithmeAi) Algorithme.Ai

Date: November 16, 2022
R Version: 4.2.1 (2019-12-12) 'Funny-Looking Kid'

Overview

This repository contains the comprehensive analytical pipeline for investigating sex-based differences in mitochondrial function and quality in pancreatic beta cells from individuals with and without type 2 diabetes (T2D). The analysis integrates single-cell RNA sequencing data with machine learning approaches to identify mitochondrial signatures associated with disease progression.

Key Features

Single-cell RNA-seq analysis of pancreatic beta cells
Sex-stratified differential expression analysis
Mitochondrial quality index (MQI) calculation and prediction using Kolmogorov-Arnold Networks (KAN)
PINK1-mediated mitophagy pathway analysis
Algorithme.AI disease scoring integration and validation
DHT treatment effects on beta cell mitochondrial function
External dataset integration (GSE217775 PINK1 KO mice)
Comprehensive visualization suite (UMAP, heatmaps, Sankey diagrams, volcano plots)

System Requirements

R Environment

R version: 4.2.1 or higher
RStudio: Recommended
Rtools: Compatible version with your R installation
Python: conda environment via reticulate

Operating System

Windows (primary development)
Linux/Mac (compatible with path adjustments)

Installation

1. CRAN Packages

install.packages(c(
  'ggplot2', 'cowplot', 'Matrix', 'ggridges', 'ggrepel', 'dplyr', 
  'plotly', 'clustree', 'patchwork', 'future', 'devtools', 'rlang', 
  'pROC', 'harmony', 'SoupX', 'tidyverse', 'viridis', 'circlize', 
  'scCustomize', 'archive', 'R.utils', 'qs', 'fastmap', 'torch',
  'readr', 'ggpubr', 'pheatmap', 'gridExtra', 'broom', 'scales', 'ggbreak'
))

2. Bioconductor Packages

if (!requireNamespace("BiocManager", quietly = TRUE))
  install.packages("BiocManager")

BiocManager::install(version = "3.18")

BiocManager::install(c(
  'BiocGenerics', 'DelayedArray', 'DelayedMatrixStats',
  'limma', 'lme4', 'S4Vectors', 'SingleCellExperiment',
  'SummarizedExperiment', 'batchelor', 'HDF5Array',
  'terra', 'ggrastr', 'EnhancedVolcano', 'DoubletFinder', 
  'glmGamPoi', 'GOSemSim', 'org.Hs.eg.db', 'AnnotationHub',
  'GenomeInfoDb', 'MeSHDbi', 'clusterProfiler', 'dittoSeq', 
  'escape', 'ComplexHeatmap', 'DropletUtils', 'Nebulosa', 
  'hdf5r', 'scDblFinder', 'JASPAR2020', 'TFBSTools', 
  'motifmatchr', 'chromVAR', 'EnrichmentBrowser',
  'BSgenome.Hsapiens.UCSC.hg38', 'EnsDb.Hsapiens.v86',
  'GenomicRanges', 'Gviz', 'rtracklayer', 'DOSE'
))

3. GitHub Packages (Version-Specific)

# Seurat ecosystem (v4.3.0.1 for Pando compatibility)
remotes::install_version(package = 'Seurat', version = '4.3.0.1')
devtools::install_version("SeuratObject", version = "4.1.4")
devtools::install_version("Signac", version = "1.11.0")

# Development versions
devtools::install_github("satijalab/seurat", ref = "develop")
devtools::install_github("satijalab/sctransform", ref = "develop", force = TRUE)

# Additional tools
devtools::install_github('cole-trapnell-lab/monocle3')
devtools::install_github("yanlinlin82/ggvenn")
devtools::install_github("gaospecial/ggVennDiagram")
remotes::install_github('chris-mcginnis-ucsf/DoubletFinder')
devtools::install_github('quadbiolab/Pando')
devtools::install_github("cole-trapnell-lab/cicero-release", ref = "monocle3")
remotes::install_github('satijalab/seurat-wrappers')

4. Verify Installation

# Check package versions
packageVersion("clusterProfiler")
packageVersion("Seurat")
packageVersion("Signac")
packageVersion("monocle3")

Data Requirements

Input Files

File	Description	Format
`processed_rna.qs`	Pre-processed Seurat object with beta cells	QS
`result.csv`	Algorithme.AI disease confidence scores	CSV
`Donor_Summary_186.csv`	Clinical metadata (HbA1c, demographics)	CSV
`AUROC.csv`	Model performance metrics	CSV
`interpretability_metrics.csv`	Model interpretability comparison	CSV
`pancreas.integrated.rds`	DHT treatment dataset (optional)	RDS
GSE217775 files	PINK1 KO mouse data (optional)	TXT

Directory Structure

Project/
├── DATA/
│   ├── QC_supplemental/
│   │   └── interpretability_metrics.csv
│   ├── DGE/
│   │   └── wilcox/
│   │       ├── by_beta_sex/
│   │       ├── cross_cluster/
│   │       ├── beta_vs_beta/
│   │       ├── T2D_vs_beta1/
│   │       └── pseudoperterbseq/
│   ├── ORA/
│   │   ├── by_beta_sex/
│   │   │   ├── UP/
│   │   │   └── DOWN/
│   │   ├── beta_vs_beta/
│   │   ├── T2D_vs_beta1/
│   │   └── ptseqbeta1_vs_t2d/
│   ├── seurat_objects/
│   │   ├── beta_cells.qs
│   │   └── combined.qs
│   ├── KANN/
│   │   ├── qc_images/
│   │   ├── fit_all_state.pt
│   │   └── input_scaler.qs
│   ├── PINK1_discovery/
│   │   └── result_with_all_features.csv
│   ├── Dataset_GSE217775/
│   │   ├── GSE217775_RAW/
│   │   ├── dge/
│   │   └── ora/
│   ├── AUROC.csv
│   ├── result.csv
│   └── Donor_Summary_186.csv
├── images/
└── REnvironments/
    └── Analytical_Pipeline.R

Analysis Workflow

1. Data Loading and Preprocessing

# Load Seurat object
processed_rna <- qread("path/to/processed_rna.qs")

# Load Algorithme.AI scores
df <- read_csv("path/to/result.csv")

# Subset and merge
subset_rna <- subset(processed_rna, cells = df$barcode)
# Add metadata
for (col in colnames(df)) {
  subset_rna[[col]] <- df[[col]]
}

2. Beta Cell Subtyping

Four beta cell subtypes identified (β1-β4):

β1: Healthy/functional beta cells (baseline)
β2: Intermediate dysfunction
β3: Stressed beta cells
β4: Most dysfunctional

# Recluster beta cells
beta_cells <- NormalizeData(beta_cells)
beta_cells <- FindVariableFeatures(beta_cells)
beta_cells <- ScaleData(beta_cells)
beta_cells <- RunPCA(beta_cells, verbose = FALSE)
beta_cells <- RunUMAP(beta_cells, dims = 1:8)
beta_cells <- FindNeighbors(beta_cells, dims = 1:8)
beta_cells <- FindClusters(beta_cells, resolution = 0.2)

# Collapse clusters into subtypes
beta_cells$collapsed_cluster <- case_when(
  beta_cells$seurat_clusters %in% c(0, 6) ~ "A",  # β1
  beta_cells$seurat_clusters %in% c(1, 5) ~ "B",  # β3
  beta_cells$seurat_clusters %in% c(2, 4) ~ "C",  # β4
  beta_cells$seurat_clusters == 3 ~ "D"           # β2
)

3. Algorithme.AI Validation

Performance Metrics:

ROC curve analysis
Confusion matrices (stratified by sex and cell type)
Correlation with HbA1c
AUROC comparison with baseline models (Random Forest, Gradient Boosting)

# ROC analysis
library(pROC)
roc_obj <- roc(subset_rna$is_T2D, subset_rna$algorithme_confidence)
plot(roc_obj, print.auc = TRUE)

# Optimal threshold
opt_thresh <- coords(roc_obj, "best", ret = "threshold")

4. Differential Gene Expression (DGE)

Three comparison strategies:

A. Within-Status (ND or T2D)

Compare β2-4 vs β1 within healthy or diseased state

# Example: β2 vs β1 in ND, stratified by sex
for (sex in c("M", "F")) {
  results <- FindMarkers(
    object = beta_cells,
    ident.1 = paste0("β2_", sex, "_ND"),
    ident.2 = paste0("β1_", sex, "_ND"),
    test.use = "wilcox",
    min.pct = 0.1,
    logfc.threshold = 0.137504  # ~10% change
  )
}

B. Cross-Status

Compare T2D β-clusters vs ND β1 (reference)

# β2 T2D vs β1 ND
for (sex in c("M", "F")) {
  results <- FindMarkers(
    object = beta_cells,
    ident.1 = paste0("β2_T2D_", sex),
    ident.2 = paste0("β1_ND_", sex),
    test.use = "wilcox"
  )
}

C. Disease Effect (T2D vs ND)

Within each β-subtype

# β2 T2D vs β2 ND
results <- FindMarkers(
  object = beta_cells,
  ident.1 = "β2_T2D",
  ident.2 = "β2_ND"
)

Test Parameters:

Test: Wilcoxon rank-sum
min.pct: 0.1 (10% minimum expression)
logfc.threshold: 0.137504 (natural log, ~10% fold-change)
Multiple testing: Bonferroni correction

5. Over-Representation Analysis (ORA)

library(gprofiler2)

# Separate UP and DOWN genes
up_genes <- filter(deg_results, p_val_adj < 0.05, avg_log2FC > 0)$gene
down_genes <- filter(deg_results, p_val_adj < 0.05, avg_log2FC < 0)$gene

# Run GO enrichment
go_results <- gost(
  up_genes, 
  organism = "hsapiens",
  correction_method = "fdr",
  sources = "GO"
)

Automated ORA Pipeline:

Processes all DGE results
Splits by UP/DOWN regulation
FDR < 0.05 threshold
Removes redundant columns
Saves to organized directory structure

6. Module Score Analysis

Pathway gene sets:

gene_sets <- list(
  UPR = c("HSPA5", "XBP1", "ATF4", "ATF6", "DDIT3", ...),
  Mitophagy = c("PINK1", "PRKN", "BNIP3", "SQSTM1", ...),
  OXPHOS = c("NDUFA9", "COX4I1", "ATP5A1", ...),
  Stress_Death = c("BAX", "BCL2", "CASP3", "TP53", ...),
  Inflammation = c("NFKB1", "TNF", "IL1B", "CXCL10", ...)
)

# Calculate scores
for (name in names(gene_sets)) {
  beta_cells <- AddModuleScore(
    beta_cells, 
    features = list(gene_sets[[name]]),
    name = name
  )
}

Correlation with disease score:

# Per-donor aggregation
df_long <- beta_cells@meta.data %>%
  group_by(Library, Sex, diabetes_status) %>%
  summarise(mean_ModuleScore = mean(Module1, na.rm = TRUE))

# Correlation test
corr_tbl <- df_long %>%
  group_by(Sex) %>%
  summarise(
    r = cor(mean_DiseaseScore, mean_ModuleScore),
    p = cor.test(mean_DiseaseScore, mean_ModuleScore)$p.value
  )

7. PINK1-Centered Analysis

A. Mitophagy Gene Expression

40+ mitophagy genes analyzed:

Core machinery: PINK1, PRKN, SQSTM1, OPTN
Receptors: BNIP3, BNIP3L, FUNDC1, PHB2
Regulators: ULK1, TBK1, MFN2, FIS1
Autophagy: MAP1LC3A/B/C, ATG5, ATG7, BECN1

mitophagy_genes <- c(
  "PINK1", "PRKN", "BNIP3", "BNIP3L", "FUNDC1", ...
)

# DotPlot visualization
DotPlot(
  beta_cells,
  features = mitophagy_genes,
  group.by = "beta_sex_diab",
  split.by = "Sex"
)

B. Pseudo-Perturbation Analysis

Strategy: Compare PINK1-high vs PINK1-low cells within ND β1

# Define PINK1 status
pink1_counts <- FetchData(beta1_nd, "PINK1", slot = "counts")[,1]
beta1_nd$PINK1_status <- "Other"
beta1_nd$PINK1_status[pink1_counts >= 2] <- "PINK1_pos"
beta1_nd$PINK1_status[pink1_counts == 0] <- "PINK1_neg"

# Differential expression
de_results <- FindMarkers(
  beta1_nd,
  ident.1 = "PINK1_pos",
  ident.2 = "PINK1_neg"
)

C. Interaction Models

Identify genes that modify PINK1 effects on T2D:

# Logistic regression with interaction
fit_interaction <- function(gene) {
  formula <- as.formula(paste0("is_T2D ~ PINK1 * ", gene))
  model <- glm(formula, data = mito_algo, family = binomial)
  
  # Extract interaction term
  tidy(model) %>% filter(term == paste0("PINK1:", gene))
}

# Run for all mitophagy genes
results <- lapply(mitophagy_genes, fit_interaction) %>%
  bind_rows() %>%
  arrange(p.value)

Visualization:

Volcano plots with FDR correction
Sex-stratified analysis
Label significant interactions

8. DHT Treatment Analysis

Dataset: Male pancreatic beta cells treated with DHT (androgen)

# Load DHT dataset
dht_dataset <- readRDS("pancreas.integrated.rds")
dht_male_beta <- subset(dht_dataset, sex == "Male" & 
                        celltype %in% c("Beta INS-hi", "Beta INS-low"))

# Map to reference
anchors <- FindTransferAnchors(
  reference = beta_cells,
  query = dht_male_beta,
  normalization.method = "SCT"
)

dht_male_beta <- MapQuery(
  anchorset = anchors,
  query = dht_male_beta,
  reference = beta_cells,
  refdata = list(beta_cluster = "beta_cluster")
)

Comparisons:

EtOH (vehicle control) vs DHT[10nM]
Mitophagy module scores
Pathway ratio analysis (OXPHOS/Glycolysis/TCA)

9. GSE217775 Analysis (PINK1 KO Mice)

Dataset: Kidney tissue from PINK1 knockout mice

Wildtype vs PINK1 KO
4 months vs 24 months
2 biological replicates per group

# Read raw count files
counts_matrix <- [combined from individual sample files]

# Differential expression (limma-voom)
dge <- DGEList(counts = counts)
dge <- calcNormFactors(dge)
design <- model.matrix(~0 + group)
v <- voom(dge, design, plot = TRUE)
fit <- lmFit(v, design)

# Contrasts
contrast.matrix <- makeContrasts(
  KO_4M_vs_WT_4M = KO_4M - WT_4M,
  KO_24M_vs_WT_24M = KO_24M - WT_24M,
  levels = design
)

Key analyses:

Mitochondrial gene expression heatmaps
Pathway module scores
Metabolic ratios (OXPHOS/TCA/Glycolysis)
Age-dependent effects

10. Kolmogorov-Arnold Network (KAN) for MQI Prediction

Mitochondrial Quality Index (MQI) Formula:

MQI = Σ(good_signals) - Σ(stress_penalties)

Good signals:
  + Mitophagy activity (weight: 1.0)
  + OXPHOS capacity (weight: 1.0)
  + UPRmt (weight: 0.7)

Stress penalties:
  - TCA overactivity (weight: 0.2)
  - MT-DNA fraction (weight: 0.5)
  - OXPHOS-Glycolysis imbalance (weight: 0.5)

KAN Architecture:

# Input features: ~50-140 mitochondrial regulators
regulators <- c(
  mitophagy_core, dynamics, biogenesis, 
  uprmt_proteostasis, antioxidant, signaling_tfs
)

# Model structure
KANNet <- nn_module(
  initialize = function(in_features, hidden = 48, n_bins = 10) {
    self$layer1 <- KANLayer(in_features, hidden, n_bins)
    self$act1 <- nn_relu()
    self$layer2 <- KANLayer(hidden, 1, n_bins)
  }
)

# Training
fit <- train_kan(
  X, y, 
  epochs = 300, 
  lr = 2e-3,
  hidden = 48,
  n_bins = 10
)

Spline-based activation:

B-spline basis functions with learnable coefficients
10 knots per input dimension
Captures non-linear relationships

Model artifacts:

fit_all_state.pt: PyTorch state dict
input_scaler.qs: Feature normalization (μ, σ)
imp_all.qs: Feature importance scores

Visualization

1. UMAP Projections

# Disease score by sex and diabetes status
p <- FeaturePlot(
  beta_cells, 
  features = "algorithme_confidence",
  split.by = "Sex_Disease"
) + scale_color_viridis_c()

2. Sankey Diagrams

Interactive flow diagrams showing pathway enrichment:

make_sankey_plot <- function(
  my_comparison_colors,
  my_pathway_colors,
  my_ora_dir,
  my_pathways
) {
  # [Implementation in script]
  # Creates alluvial flow from comparisons → pathways
  # Color by comparison, size by -log10(FDR)
}

Features:

Comparison → Pathway connections
Flow width proportional to significance
Color-coded by comparison groups
Legend with dot size scale

3. DotPlots

Gene expression heatmaps:

# Sex-aware DotPlot
DotPlot(
  beta_cells,
  features = gene_list,
  group.by = "cluster_sex_diabetes",
  dot.scale = 6,
  scale = TRUE
) + 
  facet_grid(. ~ gene_category + Sex) +
  scale_fill_gradient2(low = "white", high = "red")

Features:

Dot size = % cells expressing
Color = scaled average expression
Faceted by pathway and sex

4. Volcano Plots

PINK1 interaction effects:

volcano_plot(
  results,
  title = "PINK1 × Gene Interactions",
  xlim = c(-10, 5),
  ylim = c(0, 20)
)

Features:

X-axis: Interaction coefficient (log-odds)
Y-axis: -log10(p-value)
Colors: FDR < 0.05 (red), p < 0.05 (yellow), NS (grey)
Labels: All significant genes with ggrepel

5. Heatmaps

ComplexHeatmap visualization:

# Mitophagy genes grouped by function
Heatmap(
  expr_mat,
  col = colorRamp2(c(-1.5, 0, 1.5), 
                   c("dodgerblue4", "white", "firebrick4")),
  cluster_rows = TRUE,
  split = gene_groups,
  left_annotation = row_annot
)

Features:

Z-scored expression
Hierarchical clustering within groups
Functional category annotations
Sex-aware column grouping

6. Correlation Plots

Module scores vs Disease score or HbA1c:

# Per-donor correlation
ggplot(df, aes(x = mean_HbA1c, y = mean_ModuleScore)) +
  geom_point(aes(color = Sex, shape = Disease), size = 3) +
  geom_smooth(method = "lm", aes(color = Sex)) +
  stat_cor(aes(color = Sex))

Key Outputs

Figure Files

Figure	Description	Format
UMAP projections	Disease scores by sex/diabetes	PNG/PDF
Sankey diagrams	Pathway enrichment flows	PNG/PDF
Volcano plots	PINK1 interactions	PNG/PDF
Heatmaps	Gene expression patterns	PNG/PDF
Correlation plots	Module scores vs clinical	PNG/PDF
Violin plots	Distribution comparisons	PNG/PDF
DotPlots	Multi-gene expression	PNG/PDF

Data Tables

File	Description
`*_results.csv`	DGE results per comparison
`gost_*_UP.csv`	GO enrichment (upregulated)
`gost_*_DOWN.csv`	GO enrichment (downregulated)
`AUROC.csv`	Model performance metrics
`beta_cells_metadata.csv`	Annotated cell metadata

Model Artifacts

File	Description
`fit_all_state.pt`	KAN model weights (PyTorch)
`input_scaler.qs`	Feature normalization parameters
`imp_all.qs`	Feature importance scores
`beta_cells.qs`	Annotated Seurat object
`combined.qs`	Merged beta cells + DHT dataset

Color Schemes

Beta Cell Subtypes

beta_cols <- c(
  "β1" = "#B22222",  # firebrick
  "β2" = "#DA70D6",  # orchid
  "β3" = "#FF8C00",  # darkorange
  "β4" = "#1E90FF"   # dodgerblue
)

Sex

sex_cols <- c(
  "M" = "#1976D2",  # blue
  "F" = "#C2185B"   # pink
)

Disease Status

disease_cols <- c(
  "ND"  = "grey65",
  "T2D" = "firebrick"
)

Treatment

treatment_cols <- c(
  "Untreated" = "grey65",
  "DHT[10nM]" = "firebrick4",
  "EtOH"      = "dodgerblue4"
)

Reproducibility

Set Seed

set.seed(1234)  # Used throughout analysis

# For torch models
torch::torch_manual_seed(42)

Session Info

sessionInfo()
# R version 4.2.1 (2019-12-12)
# Platform: x86_64-w64-mingw32/x64 (64-bit)
# Running under: Windows 10 x64

Package Versions

packageVersion("Seurat")        # 4.3.0.1
packageVersion("Signac")        # 1.11.0
packageVersion("clusterProfiler") # [version]
packageVersion("ComplexHeatmap") # [version]

🧠 Algorithme.ai Models

Algorithme.ai introduces a new paradigm in machine learning — lookalike-based reasoning.
The following Python modules demonstrate two model classes built around this concept.

🐍 Part I — `tulane.py` → `SnakeClassifier`

The SnakeClassifier model establishes the foundation of the lookalike rationale.
It learns by identifying patterns and extrapolating confidence from similar data points.

from tulane import (
    make_population,
    extrapolate,
    my_function,
    datapoint_to_csv,
    my_function_audit
)
from random import choice

# Generate a list of dictionaries from a CSV file
population = make_population("train.csv")

# Extrapolate confidence percentages from 'train.csv' → 'result.csv'
extrapolate("train.csv")

# Select a random element from the population
item = choice(population)

# Print the CSV of lookalikes for the chosen datapoint
print(datapoint_to_csv(item))

# Retrieve elements in favor and against the datapoint
favorable, against = my_function_audit(item)

# Compute the confidence percentage for the specific datapoint
confidence = my_function(item)
print("Confidence:", confidence)

🦢 Part II — `tulane.py` → `BlackSwanClassifier`

The BlackSwanClassifier extends the SnakeClassifier by incorporating top feature selection and feature importance weighting, enhancing predictive precision.

from tulane import (
    make_population,
    extrapolate,
    my_function,
    datapoint_to_csv,
    my_function_audit,
    extrapolate_with_top_features,
    extrapolate_with_all_features
)
from random import choice

# Generate a list of dictionaries from a CSV file
population = make_population("train.csv")

# Standard extrapolation
extrapolate("train.csv")

# Select a random element from the population
item = choice(population)

# Print the CSV of lookalikes for the chosen datapoint
print(datapoint_to_csv(item))

# Retrieve elements in favor and against the datapoint
favorable, against = my_function_audit(item)

# Compute the confidence percentage for the specific datapoint
confidence = my_function(item)
print("Confidence:", confidence)

# Perform extrapolation using top-ranked features
extrapolate_with_top_features("train.csv")

# Perform extrapolation using all available features
extrapolate_with_all_features("train.csv")

🧩 Summary

Model	Description	Key Functions
SnakeClassifier	Base model introducing lookalike-based reasoning	`make_population`, `extrapolate`, `my_function`
BlackSwanClassifier	Enhanced model with top feature selection and broader extrapolation logic	`extrapolate_with_top_features`, `extrapolate_with_all_features`

Troubleshooting

Common Issues

1. Memory Errors

# Increase future global size
options(future.globals.maxSize = 8000 * 1024^2)

2. Missing Genes

# Check gene presence before module scores
present_genes <- intersect(gene_list, rownames(seurat_obj))
if (length(present_genes) == 0) {
  warning("No genes found in dataset")
}

3. Python Configuration

# Configure reticulate
library(reticulate)
use_condaenv("r-reticulate")
py_config()

4. UTF-8 Encoding

# For β symbols in Windows
Sys.setlocale("LC_ALL", "English")

Citation

If you use this pipeline in your research, please cite:

[Paper citation to be added upon publication]

Qadir, F. et al. (2025). Sex-based analysis of mitochondrial function in 
pancreatic beta cells reveals PINK1-dependent mechanisms in type 2 diabetes.
[Journal Name]. [DOI]

Contact

Fahd Qadir (Dragonmasterx87)
FMJ Lab
Tulane University School of Medicine
Email: [Add email address]

Acknowledgments

Algorithme.AI team for disease scoring algorithm development
Human Pancreas Analysis Program (HPAP) for donor samples and data access
GEO contributors for GSE217775 PINK1 KO dataset
Barko et al. for DHT treatment dataset
Tulane HPC for computational resources

License

[Specify license - e.g., MIT, GPL-3, Apache 2.0]

Version History

Version	Date	Changes
1.0	2022-11-16	Initial pipeline development
1.1	[Date]	Added KAN implementation
1.2	[Date]	Integrated GSE217775 analysis
1.3	[Date]	Added DHT treatment mapping

Last Updated: October 13, 2025
Pipeline Version: 1.3
Status: Production

Name		Name	Last commit message	Last commit date
Algorithm.Ai/Python Models		Algorithm.Ai/Python Models
REnvironments		REnvironments
LICENSE		LICENSE
MAGMA_analyses_script.sh		MAGMA_analyses_script.sh
ORA_analysis.zip		ORA_analysis.zip
README.md		README.md

Uh oh!

Latest commit

History

Folders and files

Repository files navigation

MitochondriaAI: Sex-Based Analysis of Mitochondrial Function in Pancreatic Beta Cells

Authors

Overview

Key Features

System Requirements

R Environment

Operating System

Installation

1. CRAN Packages

2. Bioconductor Packages

3. GitHub Packages (Version-Specific)

4. Verify Installation

Data Requirements

Input Files

Directory Structure

Analysis Workflow

1. Data Loading and Preprocessing

2. Beta Cell Subtyping

3. Algorithme.AI Validation

4. Differential Gene Expression (DGE)

A. Within-Status (ND or T2D)

B. Cross-Status

C. Disease Effect (T2D vs ND)

5. Over-Representation Analysis (ORA)

6. Module Score Analysis

7. PINK1-Centered Analysis

A. Mitophagy Gene Expression

B. Pseudo-Perturbation Analysis

C. Interaction Models

8. DHT Treatment Analysis

9. GSE217775 Analysis (PINK1 KO Mice)

10. Kolmogorov-Arnold Network (KAN) for MQI Prediction

Visualization

1. UMAP Projections

2. Sankey Diagrams

3. DotPlots

4. Volcano Plots

5. Heatmaps

6. Correlation Plots

Key Outputs

Figure Files

Data Tables

Model Artifacts

Color Schemes

Beta Cell Subtypes

Sex

Disease Status

Treatment

Reproducibility

Set Seed

Session Info

Package Versions

🧠 Algorithme.ai Models

🐍 Part I — tulane.py → SnakeClassifier

🦢 Part II — tulane.py → BlackSwanClassifier

🧩 Summary

Troubleshooting

Common Issues

1. Memory Errors

2. Missing Genes

3. Python Configuration

4. UTF-8 Encoding

Citation

Contact

Acknowledgments

License

Version History

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

🐍 Part I — `tulane.py` → `SnakeClassifier`

🦢 Part II — `tulane.py` → `BlackSwanClassifier`