HTS_IL17_Psoriasis

A FAIR Nextflow workflow and Streamlit app for integrating public HTS/qHTS assay data, psoriasis omics, proteomics, protein model features, structural evidence, and grounded evidence summaries for target and hit prioritization.

Links

• GitHub repo: https://github.com/Caffeinated-Code/HTS_IL17_Psoriasis
• Published dashboard: https://caffeinated-code.github.io/HTS_IL17_Psoriasis/
• Local Streamlit app: streamlit run app/streamlit_app.py

This is an employer-neutral portfolio project. It uses public data and compact example tables only and makes conservative claims: the workflow is a screening-to-biology prioritization workflow, not a clinical efficacy model.

What This Project Demonstrates

HTS_IL17_Psoriasis models a realistic early-discovery question:

If a pathway-proximal screening assay nominates compounds or targets around ROR gamma / Th17 biology, which candidates also have psoriasis disease evidence, proteomics support, cell-type context, and interpretable protein-level features?

The project is intentionally transparent. It favors auditable scoring, provenance, and documented limitations over black-box ranking.

Follow The Project

Start here	Markdown	Browser-friendly HTML
Analysis walkthrough	docs/analysis_walkthrough.md	analysis_walkthrough.html
Project plan	PROJECT_PLAN.md	project_plan.html
Scientific review and roadmap	DIRECTOR_REVIEW.md	director_review.html
HTS/qHTS primer	docs/hts_primer.md	hts_primer.html
IL-17 psoriasis primer	docs/il17_psoriasis_primer.md	il17_psoriasis_primer.html
Proteomics primer	docs/proteomics_primer.md	proteomics_primer.html
Protein language models primer	docs/protein_language_models_primer.md	protein_language_models_primer.html
Structure prediction primer	docs/structure_prediction_primer.md	structure_prediction_primer.html
Evidence summary primer	docs/llm_evidence_primer.md	llm_evidence_primer.html
FAIR Nextflow and AWS primer	docs/fair_nextflow_aws_primer.md	fair_nextflow_aws_primer.html

Why Psoriasis And IL-17 Biology?

Plaque psoriasis is a strong public-data case study because the IL-23 / Th17 / IL-17 axis is well established, skin biopsy datasets are available, and disease activity can be connected to immune activation, keratinocyte response, and inflammatory proteomics.

The HTS component uses public ROR gamma qHTS-style assay data as a pathway-proximal screening analog. ROR gamma t is upstream of Th17 differentiation and IL-17 production. This is not a direct IL-17 peptide screen, and the workflow calls out that limitation.

Quickstart

Requirements:

• Nextflow
• Python 3
• Streamlit, pandas, and plotly for the app

Run the local demo:

nextflow run main.nf -profile test

Launch the app after the workflow completes:

streamlit run app/streamlit_app.py

The app reads generated outputs from results/app_data/. If those files are missing, it falls back to bundled demo tables in demo_data/.

Workflow Overview

flowchart LR
    A["PubChem/ChEMBL HTS-style assay data"] --> B["Screen QC and counterscreen filtering"]
    C["Psoriasis transcriptomics"] --> D["Disease omics evidence"]
    E["Psoriasis proteomics"] --> F["Protein-level validation"]
    G["Single-cell context"] --> H["Cell-type annotation"]
    I["UniProt sequences"] --> J["ESM/ProtBERT-style features"]
    K["AlphaFold/ESMFold optional structures"] --> L["Structure confidence features"]
    B --> M["Transparent candidate ranking"]
    D --> M
    F --> M
    H --> M
    J --> M
    L --> M
    M --> N["Grounded evidence cards"]
    N --> O["HTML report and Streamlit app"]

Loading

Analysis Walkthrough

The dashboard and workflow should be read as a prioritization exercise, not an efficacy claim:

1. Start with public qHTS-style screening evidence around ROR gamma / Th17 biology.
2. Penalize assay artifacts using counterscreen logic.
3. Add psoriasis disease transcriptomics to check tissue relevance.
4. Add proteomics to test whether RNA-supported candidates also have protein-level support.
5. Add single-cell context to identify relevant immune or skin-cell compartments.
6. Add protein model and structure features as supporting interpretation layers.
7. Generate grounded evidence cards that explain the ranking, limitations, and next experiment.

Full walkthrough: docs/analysis_walkthrough.md or HTML version.

Outputs

• results/tables/candidate_rankings.tsv
• results/tables/screen_qc.tsv
• results/tables/disease_omics.tsv
• results/tables/proteomics_validation.tsv
• results/tables/singlecell_context.tsv
• results/tables/protein_features.tsv
• results/tables/structure_features.tsv
• results/evidence_cards/evidence_cards.md
• results/reports/HTS_IL17_Psoriasis_report.html
• results/provenance/run_provenance.json
• results/app_data/ tables used by Streamlit

Public Dataset Plan

The demo ships with compact cached example tables shaped like the public sources below. Full retrieval modules are documented as future work.

Evidence layer	Public source	Why it is used
HTS/qHTS	PubChem AID 2604 ROR gamma transcriptional activity screen; PubChem AID 2546 VP16 counterscreen	Screening-style pathway-proximal assay data for Th17/IL-17 biology
Bioactivity context	ChEMBL ROR gamma / IL-17 pathway assay records	Curated drug discovery assay context
Disease transcriptomics	GSE54456 psoriasis lesional vs normal skin RNA-seq	Disease expression support
Proteomics	PRIDE PXD021673 psoriasis skin LC-MS/MS	Protein-level validation
Single-cell	GSE162183 psoriasis skin scRNA-seq	Cell-type context

Director-Level Caveats

• ROR gamma qHTS is pathway-proximal: it is relevant to Th17 / IL-17 biology but is not a direct IL-17 peptide screen.
• Public screening data are small-molecule assays, used here as an HTS data analog.
• Protein language model features are descriptive unless validated in a predictive task.
• AlphaFold or ESMFold confidence supports structural plausibility, not binding proof.
• Evidence summaries are grounded in workflow outputs and should never create unsupported biological claims.

For a critical staff-scientist-style review and sequential improvement roadmap, see DIRECTOR_REVIEW.md.

Documentation Primers

• HTS primer / HTML
• IL-17 psoriasis primer / HTML
• Proteomics primer / HTML
• Protein language models primer / HTML
• Structure prediction primer / HTML
• Evidence summary primer / HTML
• FAIR Nextflow and AWS primer / HTML