JCaperella Bioinformatics Hub

JC Enrichment Network Studio (Dash App)

Interactive Dash web app for exploring gene ↔ pathway/term enrichment results as a bipartite network. Upload long-format enrichment memberships, filter and explore hubs, view summary stats, and export node/edge tables for downstream analysis (e.g., Cytoscape).

• 📄 Input: long-format CSV (one row per gene–term membership edge)
• 🧬 Bipartite graph: genes ↔ pathways/terms
• 🎚️ Filters: search, min degree, min edge weight, max groups, layout mode
• 📊 Stats: node/edge counts, components, top genes/terms by degree
• ⚖️ Weighted edges: supports padj/FDR; auto-converts to -log10(p) for plotting
• ⬇️ Export: download nodes.csv and edges.csv
• ☁️ Live deployment: Google Cloud Run

Tip: If you have a standard enrichment table (term + list of genes), convert it to long format first (one gene per row per term). This app is designed for clean network building and fast exploration.

Bulk RNA-Seq Analyzer

Interactive Shiny app for bulk RNA-seq: differential expression, PCA/UMAP, volcano & heatmaps, enrichment, Random Forest, power analysis, and downloadable results.

• 🗂️ Inputs: counts matrix + phenotype CSV
• 🧬 DE: limma-voom workflow
• 🧭 Dims: PCA & UMAP
• 🌋 Plots: volcano, heatmap, interactive tables
• 🧠 ML: Random Forest + ROC/AUC
• 🧪 Pathways: Enrichr (KEGG/GO/Reactome)
• ⚡ Power: sample size/power curves
• designed for HPC environments using Singularity

Paired FASTQ QC (WDL + Cromwell)

Containerized WDL workflow executed with Cromwell for paired-end FASTQ QC. Runs FastQC on R1/R2 and generates a single MultiQC report across samples, plus a merged read-count table.

• 🧬 Inputs: paired FASTQs (R1/R2) + sample IDs via JSON
• 🔍 QC: FastQC per read pair
• 📊 Multi-sample summary: MultiQC report
• 🧾 Counts: merged per-sample read counts (R1/R2)
• 📦 Reproducible: Docker runtime inside WDL tasks

Use case: a compact, portfolio-friendly example of WDL/Cromwell workflow wiring, containerized execution, and QC aggregation across multiple samples.

Counts_matrix_Nextflow (RNA-seq Pipeline)

Portable Nextflow DSL2 RNA-seq workflow: QC → STAR alignment → gene counts → Salmon TPM → BigWig coverage → matrix merge + MultiQC summary. Designed to be minimal, readable, and robust.

• 📄 Inputs: paired-end FASTQs via samplesheet CSV
• 🧬 Core steps: fastp → STAR → featureCounts → Salmon → MultiQC
• 📈 Outputs: gene counts matrix, TPM matrix, BigWig tracks, QC report
• 🧱 Indexes: auto-builds STAR + Salmon indexes (per run)
• 🛡️ Robustness: skips coverage gracefully for zero-mapped samples
• 📦 Runs anywhere: Conda, Docker, or Singularity/Apptainer
• 🧪 Great for: demos, infra tests, portfolio/template pipelines

Quickstart

Docker

docker build -t rnaseq-pipeline .
nextflow run main.nf -profile docker \
  --samplesheet samples.csv \
  --ref genome.fa \
  --gtf genes.gtf \
  --transcripts transcripts.fa
    

Singularity/Apptainer

singularity build containers/rnaseq-pipeline.sif docker://rnaseq-pipeline
nextflow run main.nf -profile singularity \
  --samplesheet samples.csv \
  --ref genome.fa \
  --gtf genes.gtf \
  --transcripts transcripts.fa
    

Key Outputs

results/
├── qc/                       (fastp)
├── ref/                      (STAR + Salmon indexes)
├── bam/                      (sorted BAM + BAI + flagstat)
├── bigwig/                   (coverage tracks)
├── counts_per_sample/        (per-sample featureCounts)
├── counts_matrix.tsv         (gene count matrix)
├── salmon_tpm_matrix.tsv     (transcript TPM matrix)
└── multiqc_report.html       (summary report)
    

Notes: intentionally simple; explicit channels; guards against common failure modes. Easy to extend into DESeq2/edgeR downstream analysis.

Enrichment Analysis LLM Triage (Flask App)

A lightweight Flask web app that takes enrichment results (CSV) and generates a structured, human-readable triage report — highlighting likely biological drivers, reactive programs, potential confounders, and suggested follow-up experiments.

• 📄 Input: enrichment CSV (terms + scores + genes/overlap fields)
• 🧠 LLM reasoning: summarizes key programs and flags confounding patterns
• 🧪 Follow-ups: proposes targeted experiments with readouts + controls
• 📑 PDF report: generates a clean downloadable triage PDF
• 📦 Deploy: Docker + Apptainer/Singularity (HPC-friendly)

Why it matters: enrichment tables are easy to generate but hard to interpret. This tool helps translate “significant pathways” into an actionable short list of mechanisms and concrete next experiments — without drowning the user in jargon.

Breast Cancer NLP Phenotyper (Dash + medspaCy)

A lightweight, rule-based clinical NLP dashboard for extracting key breast cancer phenotypes from free-text notes (e.g., pathology, consults). Outputs include a clean patient-level table plus auditable evidence mentions with snippets so results can be reviewed and trusted.

• 📄 Input: upload multiple .txt notes + optional mapping CSV (note → patient/date/type)
• 🧠 NLP engine: spaCy + medspaCy patterns (deterministic extraction)
• 🧬 Phenotypes: ER/PR status (+ % if present), HER2 (IHC/FISH → final), Ki-67, histology, grade, stage
• 🔎 Evidence: each extracted value is backed by mention-level snippets for auditing
• 🧮 Aggregation: note-type/date precedence rules roll note-level data to patient-level output
• designed for HPC environments using Singularity

Note: This is an MVP designed for transparency and portability — ideal for demos, iteration, and extension into richer rule sets or model-assisted extraction later.

ATAC-Seq Peak Annotation & Enrichment

Upload MACS2 .narrowPeak, annotate with ChIPseeker, and run GO/KEGG/Reactome enrichment with slick visuals and CSV exports.

• 📄 Input: MACS2 .narrowPeak file
• 🏷️ Annotation: ChIPseeker + TxDb
• 📊 Views: pie charts, tables, barplots
• 🧠 Pathways: enrichR (GO/KEGG/Reactome)
• designed for HPC environments using Singularity

miRNA Differential Expression & Enrichment

DESeq2-based miRNA analysis with PCA/UMAP, volcano & heatmaps, Enrichr enrichment, Random Forest classification, power analysis, and exports.

• 🗂️ Inputs: miRNA counts + metadata
• 🧬 DE: DESeq2 pipeline
• 🧭 Dims: PCA & UMAP
• 🌋 Plots: volcano, top-miRNA bars, heatmaps
• 🧠 ML: RF classification + metrics
• 🧪 Pathways: Enrichr (clusterProfiler fallback)
• ⚡ Power: sample size estimates
• designed for HPC environments using Singularity

DNA Methylation App

Explore beta values, run differential methylation, enrichment, PCA/UMAP, Random Forest, power analysis, and download everything — HPC-ready.

• 🗂️ Input: CSV beta matrix (e.g., 450k)
• 🧪 DE: probe-level stats + FDR
• 🧭 Dims: PCA & UMAP
• 🧠 ML: RF + AUC & importance
• 🧪 Pathways: Enrichr KEGG/GO/Reactome
• ⚡ Power: Cohen’s d → n per group
• 📦 Deploy: Singularity/Apptainer

CRISPR Mixscape Pipeline (Perturb-seq)

Single-cell CRISPR screen workflow using Seurat’s Mixscape: QC/normalization, UMAP, perturbation scoring, KO/NP/NT assignment, DE, and rich plots — with HPC support.

• 🧪 Inputs: counts + metadata CSVs
• 🧭 Dims: UMAP visualization
• 🧮 Mixscape: perturbation scores & class labels
• 🧬 DE: KO vs NT + downloads
• 📈 Views: bar/violin/heatmaps, summaries
• 📦 Deploy: Singularity + Slurm script

GWAS Analysis App

A full-stack, no-code Shiny app for Genome-Wide Association Studies (GWAS) using raw VCF files—no PLINK needed. Upload your VCF, phenotype, and covariate table to begin.

• 🧪 QC Filters: MAF, allele frequency, call rate, HWE p-value thresholds
• 🧮 GWAS Engine: Logistic regression with Bonferroni correction support
• 📊 Visualization: PCA, UMAP, QQ plot, Manhattan plot with region zoom
• 🤖 Machine Learning: Random Forest with AUC, importance, ROC
• 🧬 SNP-to-Gene Mapping: Map significant SNPs to nearest genes
• 🧠 Enrichment: KEGG/GO/Reactome via enrichR
• ⚡ Power Analysis: Cohen’s d-based observed power & sample size curve
• 📦 Export Everything: GWAS tables, enrichment results, ML metrics, more
• designed for HPC environments using Singularity

About Me

I’m John Caperella, a bioinformatics developer passionate about turning complex genomic data into usable insights. I build clear, scalable tools for single-cell RNA-seq, CRISPR Mixscape, and omics visualization in R, Python, and Shiny—helping researchers and data scientists get to answers faster.

• 🧬 Single-cell & CRISPR analytics (Seurat/Mixscape)
• 📊 Reproducible visualization apps (Docker/Singularity/HPC)
• 🤖 Exploring LLMs for genomics workflows

Contact Form LinkedIn GitHub Résumé (HTML)

Your Email Subject Message Send Message

Research Radar (PubMed + LLM Summaries)

A lightweight research intelligence dashboard that pulls recent PubMed papers across multiple queries and generates 3-bullet summaries using a local LLM (Ollama). Includes filters, trending topics, and top journals — built for fast scanning.

• 🧠 LLM summaries: exactly 3 bullets per abstract
• 📈 Dashboard: trending topics + top journals + search/filter
• 🔁 Update workflow: regenerate papers.json and push to Pages
• 🧱 Portfolio pattern: API ingestion → pipeline → structured JSON → UI

Tip: This is ideal for tracking AI-for-biology, computational methods, and variant calling literature without drowning in full abstracts.

GCP FASTQ Event Pipeline (Eventarc → Cloud Run Job)

Cloud-native, event-driven FASTQ processing on Google Cloud. Upload a FASTQ to Cloud Storage and automatically trigger a serverless pipeline: Eventarc fires on object finalize → Cloud Function launches a containerized Cloud Run Job → results are written back for downstream analysis.

• ☁️ Trigger: GCS “object finalize” event (Eventarc)
• 🧩 Orchestration: Cloud Function (glue layer)
• 📦 Compute: Cloud Run Job (containerized batch)
• 🧬 Bioinformatics-ready pattern: scalable ETL wiring for genomics ingestion
• 📤 Outputs: structured results written to storage (and designed to extend to analytics/warehousing)
• 🛠️ Great for: portfolio demos of serverless pipelines + reproducible containers

Why it matters: This mirrors real-world genomics platform patterns — automatic ingestion triggers + containerized compute, without manual job submission.

Document Cleaning CLI

AI-powered document cleanup for scanned pages, noisy screenshots, and OCR-bound records. Enhances messy source images into cleaner, sharper outputs that are easier to read, extract, and route into downstream research or clinical workflows.

• 📥 Inputs: scanned .png/.jpg files or ZIP batches of document images
• 🧼 Cleanup: deep-learning denoising + image enhancement for messy source material
• 📄 Outputs: OCR-optimized images and PDF-ready cleaned documents
• 🛠️ Modes: command-line workflow or REST API deployment
• 🏥 Use cases: legacy records, scanned notes, exported forms, and other hard-to-read document inputs
• ⚙️ designed for flexible local or API-based deployment workflows

Why it matters: a lot of valuable information is trapped inside low-quality visual documents. This tool helps turn noisy records into cleaner, machine-readable assets for OCR, review, and downstream automation.

Usage Analytics

This dashboard is powered by Google Analytics 4 and Looker Studio. It shows historical click activity for each app.

Last checked: loading…