rwe 0.1.4

Real World Evidence utilities and reporting

Real world evidence of siRNA targets

The current pipeline generates real world genetic evidence documents for an siRNA target using either cross-source summary statistics or biobank-specific participant-level data. The report can be used for the following three broader utilities:

• Discover new target-indication pairs
• Safety evaluation of potential target
• Repurposing opportunity of existing target

Description of the report

The package now supports two report modes:

• summary
• biobank

Summary report

The summary report includes only cross-biobank and source-level summary evidence:

• Clinical records
• Labs and measurements
• Biomarkers
• Indication-specific reports

It does not include:

• Variant information
• Demographics
• Labs result profile images
• Survey information
• Homozygous loss of function carriers

Biobank report

The biobank report is biobank-specific and currently supports aou.

It includes:

• Variant information and demographics
• Labs and measurement figures
• Survey information
• Homozygous loss of function carriers

Future updates might have the following additional sections:

• OpenTargets
• Knowledge portal networks: https://hugeamp.org/research.html?pageid=kpn_portals
• Genomics England information: https://www.genomicsengland.co.uk/
• Genes and Health information: https://www.genesandhealth.org/
• Generate automated ppt report
• Link AI via api to generate summary of the indication specific section
• Use AI to generate ppt report for efficacy

Variant information and demographics

Variant information

Provides number of pLoF carriers across four variant categories in the All of Us cohort:

• stop gained
• frameshift
• splice acceptor
• splice donor

Demographics

Includes age, sex, ancestry and ethnicity information of pLoF carriers in comparison with non-carriers.

Clinical records

Provides phenomewide association study results of pLoF carriers in All of Us and UK Biobank cohorts. The All of Us association results are generated in-house. The UK Biobank results are collected from GeneBass and the AstraZeneca PheWAS portal.

Labs and measurements

Provides lab and measurement association results for pLoF carriers in All of Us and UK Biobank cohorts in comparison to non-carriers. Detailed measurement definitions and concept IDs are maintained in docs/labs_and_measurements.md (included in the source distribution).

Survey information

Includes self-reported survey information about general, mental, physical and overall health of pLoF carriers in comparison with non-carriers in the All of Us cohort. This section is used only in biobank mode.

Homozygous loss of function carriers

Provides demographics and survey information of the biallelic lof variant carriers in All of Us.

Biomarkers

Provides association statistics of gene pLoF with biomarker measurements, including plasma protein levels.

Indication specific report

Provides association results for user-specified indications from All of Us and UK Biobank cohorts.

The recommended and current default workflow uses the target-profiler indication mapping database and requires an explicit SQLite database path. Traits are classified into:

• core
• established
• peripheral
• unrelated

The indication tables now include a Predicted association column, and rows are grouped in this order:

• core
• established
• peripheral
• unrelated

Within each group, rows are sorted by p-value.

Traits that do not exactly match the database trait universe are not included in the indication table. They are written to source-specific log files in the report output directory:

• aou_unmatched_indication_traits.txt
• genebass_unmatched_indication_traits.txt
• astrazeneca_unmatched_indication_traits.txt

Keyword-based indication matching remains available only as an explicit legacy option and is being phased out.

The indication scope has expanded substantially and is now documented separately in docs/indication_scope.md, including the current database-backed indication count and trait-universe size.

Resources used to generate the report

Controlled Datasets

All of Us

The All of Us cohort currently consists of 420k participants with whole genome sequencing and phenotypic data.

Open Source Databases

Here we describe the open source databases used for gathering evidence about the targets:

GeneBass

GeneBass reports phenomewide associations for LoF carriers among 380k participants from the UK Biobank cohort.

AstraZeneca PheWAS portal

AstraZeneca reports phenomewide associations for LoF carriers among 500k participants from the UK Biobank cohort.

Updates and Installation

Separately in TODO

Internal Use for installation

# upgrade packages for building
python -m pip install -U pip build
pip install twine

# New version packaging and upload
rm -rf dist build *.egg-info src/*.egg-info
# PowerShell:
Remove-Item -Recurse -Force dist, build, *.egg-info, src\*.egg-info
conda activate rwe
python -m build
pip install dist/rwe-0.1.4-py3-none-any.whl
python -c "from rwe.generate_report import generate_rwe_report; import rwe.clients.aou as aou; import rwe.clients.azn as azn; import rwe.clients.genebass as gbs; print('import ok')"
twine upload dist/*

# Before packaging environment test
conda install -c conda-forge python=3.12
pip install -r requirements.txt
playwright install
python -m playwright install-deps

Report configuration

The report generator supports two explicit modes:

• summary
• biobank

Example:

from rwe.generate_report import generate_rwe_report

generate_rwe_report(
    gene="PCSK9",
    chrm="1",
    mode="summary",
    indications="hyperlipidemia",
    indication_mapping_mode="database",
    indication_db_path=r"C:\Users\dbanerjee\work\target-profiler\data\mappings\database\indication_trait_gpt5.sqlite",
    aou_project_dir="s3://your-bucket/summary_statistics/plof/processed",
    allofus=True,
    genebass=True,
    astrazeneca=True,
    verbose=True,
)

Set verbose=True to print the indication-slide conclusion to stdout while the full run log is still written to the .log file next to the DOCX output.

Summary mode with a user-defined AoU project directory:

generate_rwe_report(
    gene="NPPA",
    chrm="1",
    mode="summary",
    indications="heart_failure",
    indication_mapping_mode="database",
    indication_db_path=r"C:\Users\dbanerjee\work\target-profiler\data\mappings\database\indication_trait_gpt5.sqlite",
    aou_project_dir="s3://your-bucket/summary_statistics/plof/processed",
    allofus=True,
    genebass=True,
    astrazeneca=True,
    out_docx_path="data/NPPA/RWE_NPPA_summary_report.docx",
)

Multiple genes and indications in summary mode:

from pathlib import Path

genes = [
    {"gene": "MC4R", "chrm": "18", "indications": "obesity"},
    {"gene": "PCSK9", "chrm": "1", "indications": "hyperlipidemia"},
]

root = Path("data/batch_summary_runs")
root.mkdir(parents=True, exist_ok=True)

for item in genes:
    gene_dir = root / item["gene"]
    gene_dir.mkdir(parents=True, exist_ok=True)
    generate_rwe_report(
        gene=item["gene"],
        chrm=item["chrm"],
        mode="summary",
        indications=item["indications"],
        indication_mapping_mode="database",
        indication_db_path=r"C:\Users\dbanerjee\work\target-profiler\data\mappings\database\indication_trait_gpt5.sqlite",
        aou_project_dir="s3://your-bucket/summary_statistics/plof/processed",
        allofus=True,
        genebass=True,
        astrazeneca=True,
        out_docx_path=str(gene_dir / f"RWE_{item['gene']}_summary_report.docx"),
        generate_pptx=True,
        verbose=True,
    )

Legacy keyword mode example:

generate_rwe_report(
    gene="PCSK9",
    chrm="1",
    mode="summary",
    indications="obesity",
    indication_mapping_mode="keywords",
    allofus=True,
    genebass=True,
    astrazeneca=True,
)

Biobank example:

generate_rwe_report(
    gene="PCSK9",
    chrm="1",
    mode="biobank",
    biobank="aou",
    generate_pptx=False,
)

Resources

1. ICD to Phecode mappings: https://www.vumc.org/wei-lab/sites/default/files/public_files/ICD_to_Phecode_mapping.csv
2. gnomAD v4.1 constriant metrics: https://gnomad.broadinstitute.org/data
3. phecodeX labels: https://github.com/PheWAS/PhecodeX
4. nptv carriers gnomad and genebass: Internal (Shicheng)
5. nptv carriers aou: Internal (Deepro)
6. Clingen haploinsufficiency curation: https://search.clinicalgenome.org/kb/downloads#section_dosage
7. Decipher haploinsufficiency index: https://www.deciphergenomics.org/files/downloads/HI_Predictions_Version3.bed.gz
8. Clingen disease summary: https://search.clinicalgenome.org/kb/downloads#section_gene-disease-validity