Psytar

Skill BioTender-max/awesome-bio-agent-skills/skills/drugclaw/psytar

A curated collection of AI agent skills for biomedical research, covering genomics, proteomics, single-cell analysis, clinical AI, and protein design.

Install

npx -y skills add BioTender-max/awesome-bio-agent-skills --skill psytar

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.

What its author says it does

Copied from the file, not written here

Query the PsyTAR psychiatric adverse-reaction corpus. Use when the user asks about patient-reported ADRs, withdrawal symptoms, drug indications, or effectiveness for Zoloft, Lexapro, Cymbalta, or Effexor XR. Accepts drug names (brand or generic), symptom terms, or UMLS CUIs.

SKILL.md

4.0 KB, as published. Nobody here has run it

PsyTAR Query Skill

891 patient reviews → 6 009 annotated sentences → extracted ADR / WD / SSI / DI entities → mapped to 918 UMLS + 755 SNOMED CT concepts.

Entity Detection & Routing

Input Pattern	Detected As	Targets
`C0917801`	UMLS CUI	`*_Mapped` sheets only (UMLS1/UMLS2 cols)
`Zoloft` / `sertraline`	Drug name	`drug_id` or `drug` columns (alias-aware)
`nausea`, `insomnia` …	Free text	substring across all cell values

Generic ↔ brand aliases: sertraline↔Zoloft, escitalopram↔Lexapro, duloxetine↔Cymbalta, venlafaxine↔Effexor.

API

Function	Input	Returns
`search(entity, sheet?, label?)`	single string	`{sheet: [row_dict]}`
`search_batch(entities, sheet?, label?)`	list of strings	`{entity: {sheet: [row_dict]}}`
`summarize(results, entity)`	result dict + label	compact text
`to_json(results)`	result dict	`list[dict]` (flat, adds `_sheet`)
`describe()`	—	dataset overview text

Parameters

sheet — restrict to one sheet (fuzzy-matched: "ADR" → ADR_Identified).
label — when on Sentence_Labeling, keep only rows where the named label column (ADR / WD / EF / INF / SSI / DI) equals 1.

Sheet Schema

Sheet	Description	Key Columns
`Sample`	Original posts	drug_id, rating, indication, side-effect, comment, gender, age, duration
`Sentence_Labeling`	6 009 sentences, binary labels	drug_id, sentence_index, sentences, ADR, WD, EF, INF, SSI, DI, Findings, others, rating, category
`ADR_Identified`	Extracted ADR mentions	drug_id, sentence_index, sentences, ADR1 … ADRn
`WD_Identified`	Extracted WD mentions	drug_id, sentence_index, sentences, WD1 … WDn
`SSI_Identified`	Extracted SSI mentions	drug_id, sentence_index, sentences, SSI1 … SSIn
`DI_Identified`	Extracted DI mentions	drug_id, sentence_index, sentences, DI1 … DIn
`ADR_Mapped`	ADR → UMLS/SNOMED	drug_id, sentence_index, ADR/ADRs, UMLS1, UMLS2, SNOMED-CT, mild, moderate, severe, persistent, not-persistent, body-site, rating, drug, class, type, entity_type
`WD_Mapped`	WD → UMLS/SNOMED	(same structure as ADR_Mapped)
`SSI_Mapped`	SSI → UMLS/SNOMED	(same structure)
`DI_Mapped`	DI → UMLS/SNOMED	(same structure)

Mapped-sheet qualifier columns

mild, moderate, severe — severity descriptors; persistent, not-persistent — duration; body-site — anatomical site; entity_type — Cognitive / Physiological / Psychological / Functional.

Usage

from importlib.machinery import SourceFileLoader
m = SourceFileLoader("psytar", "36_PSYTAR.py").load_module()

# overview
print(m.describe())

# drug → ADR mappings
res = m.search("Zoloft", sheet="ADR_Mapped")
print(m.summarize(res, "Zoloft"))

# generic name works too
res = m.search("sertraline", sheet="ADR_Mapped")

# symptom in one Identified sheet
res = m.search("nausea", sheet="ADR_Identified")

# symptom across all sheets
res = m.search("insomnia")

# UMLS CUI (auto-scoped to Mapped sheets)
res = m.search("C0917801")

# withdrawal sentences for Effexor
res = m.search("Effexor", sheet="Sentence_Labeling", label="WD")

# batch
batch = m.search_batch(["Lexapro", "insomnia", "C0917801"])

# JSON for pipeline
flat = m.to_json(m.search("Cymbalta"))

Data Source

Corpus: PsyTAR v1.0 — CC BY 4.0
File: PsyTAR_dataset.xlsx — set via DATA_PATH or env PSYTAR_XLSX
Paper: Zolnoori et al., Data in Brief 24, 103838 (2019). https://doi.org/10.1016/j.dib.2019.103838
Stats: 891 reviews, 6 009 sentences, 4 813 ADR + 590 WD + 1 219 SSI
- 792 DI mentions, 918 UMLS / 755 SNOMED concepts