Scientific Research Assistant

Problem

Research workflows often split across tools: databases for literature, Python for statistics, plotting libraries for figures, manual reference cleanup for papers, and separate compound databases or docking tools for drug discovery. Scientific Research Assistant connects this chain to an Agent that supports SKILL.md and exec, so researchers can start a task with natural language or structured parameters and inspect intermediate artifacts under output/.

Core Capabilities and Key Steps

• Literature review: search PubMed, Google Scholar, and arXiv by topic; trace forward and backward citations; define inclusion and exclusion criteria; assess bias and quality; structure background, methods, results, and discussion; and handle APA, Nature, and IEEE citation formats.
• Data analysis and visualization: cover descriptive statistics, t-test, ANOVA, regression, PCA, clustering, Kaplan-Meier, and Cox models; run bioinformatics flows such as BLAST, Scanpy, and KEGG; and generate heatmaps, volcano plots, UMAP, and t-SNE visualizations.
• Drug discovery: support target identification, virtual screening, and SAR optimization using tools such as Open Targets, ChEMBL, PubChem, ZINC, AutoDock, and RDKit.
• Paper and grant support: organize manuscripts using IMRaD; prepare abstracts, figure legends, Cover Letter, and journal formatting; and produce grant-ready hypotheses, project plans, budgets, and impact statements.

Boundaries

This skill is best used as engineering-level and semi-automated research assistance. It does not replace scientific judgment, peer review, ethics review, clinical diagnosis, or wet-lab operations. For heavy computation, external databases, or cloud GPUs, verify network access, dependency versions, random seeds, API rate limits, and secret handling through environment variables rather than hardcoded credentials.