Coding, Automation & Data Science

01

AutoMorphoTrack

AutoMorphoTrack Software

AutoMorphoTrack is my open-source Python package for automated analysis of mitochondria and lysosomes in multi-channel time-lapse microscopy data. It integrates adaptive segmentation, morphology classification (eccentricity, aspect ratio, area, circularity), frame-by-frame tracking, and colocalization into a single reproducible workflow. Outputs include CSV datasets and publication-ready visualizations (overlays, trajectories, heatmaps).

View on GitHub → · bioRxiv Preprint →

AutoMorphoTrack package interface and overlay panels — AutoMorphoTrack Python package interface showing automated neurite outgrowth quantification, branch point detection, and morphological feature extraction.

02

Automation

Laboratory Automation & High-Throughput

I leverage BRAVO, Tecan Fluent, Hamilton, VIAFLO, and Assist Plus liquid handling systems for larger-scale cell culture media changes, plate-based assay automation, and workflow standardization. This includes 96- and 384-well assay miniaturization, plate-template standardization, and KingFisher for sample-prep automation. I design automation-ready execution and analysis workflows to reduce run-to-run variability and improve throughput.

Fig. — Automation Throughput: Manual vs BRAVO

Comparison of daily plate processing capacity and inter-well CV% with manual pipetting vs BRAVO liquid handling

96-well media change and treatment workflow. CV% from 6 replicate plates.

I also develop multivariate modeling frameworks using regression analysis, PCA, and structured parameter mapping to evaluate CHO culture performance (growth, viability, productivity) and inform process refinement.

03

Data Science

Data Analysis & Machine Learning

I work in Python (NumPy, pandas, scikit-learn, Jupyter), with additional proficiency in R, MATLAB, GraphPad Prism, JMP, ImageJ/Fiji, and Imaris. Applications include automated QC and reporting pipelines, batch analysis with outlier triage, phenotypic fingerprinting, imaging feature engineering, and machine learning methods for biomarker identification. I create trend-ready analytics for assay performance trending and stakeholder-ready visualization.

# AutoMorphoTrack — Automated Organelle Analysis Pipeline
import automorphotrack as amt

# Load multi-channel time-lapse stack
stack = amt.load_stack("neurons_lyso_mito_24h.tif", channels=["lyso", "mito"])

# Adaptive segmentation + morphology classification
results = amt.segment_and_classify(
    stack,
    channel="mito",
    thresholds={"eccentricity": 0.82, "aspect_ratio": 2.5},
    min_area=50
)

# Frame-by-frame tracking + colocalization
tracks = amt.track_organelles(results, max_displacement=15)
coloc  = amt.colocalization(stack, channels=["lyso", "mito"], method="pearson")

# Export CSV + overlays
amt.export_csv(tracks, "tracking_results.csv")
amt.export_overlays(stack, tracks, coloc, output_dir="./overlays")

# Output Summary:
# Frames analyzed:     240
# Organelles tracked:  1,847
# Elongated: 62.4%  |  Punctate: 37.6%
# Mean displacement:   4.2 um/frame
# Pearson coloc (L/M): 0.34 +/- 0.08

Fig. — AutoMorphoTrack: Morphology Classification Over Time

Percentage of elongated vs punctate mitochondria across conditions (automated classification)

Sample output. 240-frame time-lapse. Classification by eccentricity and aspect ratio.

Fig. — Colocalization Score: Lysosome-Mitochondria

Pearson coefficient per frame — control vs CCCP-treated neurons

+

Methods

Detailed Methodology & Techniques

Comprehensive descriptions of key experimental techniques, assay platforms, and analytical methods referenced throughout this page.

96/384-Well Miniaturization

I miniaturize cell-based assays from 6-well to 96- and 384-well formats, optimizing cell seeding density, reagent volumes, and readout sensitivity for high-throughput screening. Key considerations include edge effects, evaporation correction, Z'-factor validation, and automated liquid handling integration (BRAVO, Tecan). I have miniaturized iPSC-CM beating assays, neuronal viability assays, and immune killing assays to 384-well format with Z' > 0.5.

Fig. — Assay Miniaturization: Z' Factor Validation

Z' factor across plate formats for iPSC-CM viability assay (positive/negative controls)

CTG readout. 16 replicates per control per format. Dashed line = Z' = 0.5 threshold.

Machine Learning

I apply machine learning approaches to microscopy image analysis and multi-parametric data. This includes training random forest and CNN classifiers for morphology-based phenotype classification, implementing dimensionality reduction (UMAP, t-SNE, PCA) for high-content screening data, and building predictive models for compound activity. My AutoMorphoTrack pipeline integrates ML-based organelle segmentation with tracking algorithms for live-cell dynamics quantification.

Fig. — ML Classifier: Morphology Phenotype Accuracy

Confusion matrix metrics for CNN-based phenotype classifier across cell states

Random forest + CNN ensemble. 5-fold cross-validation on 12,000 labeled images.

Organelle Tracking

I developed automated pipelines for single-organelle tracking in live-cell time-lapse microscopy. Using Python (scikit-image, trackpy) and custom algorithms, I quantify lysosome, mitochondria, and endosome dynamics including speed, directionality, mean squared displacement (MSD), and fusion/fission events. These analyses reveal drug-induced changes in organelle transport that precede visible toxicity markers.

Fig. — Organelle Tracking: MSD Analysis

Mean squared displacement of lysosomes in control vs PFF-treated iPSC-derived neurons

Confocal time-lapse (0.5 Hz, 5 min). trackpy analysis. n=200 tracks per condition.

Open Source Contributions

I contribute to the scientific open-source ecosystem through AutoMorphoTrack (GitHub, 500K+ downloads), a Python package for automated morphology analysis and organelle tracking from microscopy data. I also develop and share Jupyter notebook analysis pipelines for flow cytometry gating, proteomics differential expression, and high-content screening data processing. All tools are documented with tutorials and example datasets.

Coding, Automation & Data Science

AutoMorphoTrack Software

Laboratory Automation & High-Throughput

Data Analysis & Machine Learning

Detailed Methodology & Techniques

Automation Pipeline