Methods & Data Sources

InChemico.info is a real-time biochemical analysis platform that queries public scientific databases to surface gene-disease relationships, drug interactions, natural product evidence, and toxin exposures. All data is retrieved live from authoritative sources — no proprietary database is maintained.

The platform integrates five public databases: PubMed/NCBI (biomedical literature, 35M+ citations), NCBI Gene (gene records, orthologs, genomic coordinates), PubChem (small molecule structures, bioassay data, 115M+ compounds), OpenFDA (adverse event reports, drug labels, recalls), and DailyMed (FDA-approved drug labeling, mechanism of action, pharmacogenomics).

Queries are submitted to NCBI E-utilities (esearch, efetch, elink, esummary) using MeSH terms and free-text Boolean operators. PubChem is queried via the PUG REST API. OpenFDA and DailyMed are queried via their respective REST APIs. All outbound requests use exponential backoff with jitter and respect each API's rate limits (NCBI: 3 req/s without API key; PubChem: 5 req/s; OpenFDA: 240 req/min).

Search results are ranked by a composite confidence score combining: citation count (log-normalized), recency (exponential decay over 10 years), evidence tier (clinical trial > systematic review > case report), and source diversity (multi-database corroboration). Scores are normalized to a 0–100 scale. Scores above 70 indicate strong multi-source evidence; 40–70 moderate evidence; below 40 limited or preliminary evidence.

The SBIR Analyzer uses rule-based heuristic checks calibrated against 7 scored NIH applications (impact scores 16–47) and 10 cross-cutting weakness patterns identified from NIH study section feedback. Checks cover: hypothesis presence, go/no-go criteria, statistical plan, pitfalls/alternatives, writing quality (passive voice density, jargon, paragraph length), criteria alignment (Significance, Innovation, Approach, Investigator, Commercialization), and post-reauthorization compliance rules. Estimated scores are heuristic approximations — actual reviewer scores depend on study section, payline, and full application context.

Protein and nucleotide sequences are retrieved from NCBI via elink (gene-to-protein) and displayed using a ClustalX-style residue coloring scheme. Conservation tracks are computed from pairwise identity. Full BLAST integration for homolog discovery is in development.

InChemico.info is a research tool for hypothesis generation, not clinical decision-making. Evidence is retrieved from public databases and may be incomplete, outdated, or contain errors. Drug interaction data from OpenFDA reflects reported adverse events, not established causal relationships. SBIR analysis is heuristic and does not substitute for expert grant review. All results should be independently verified against primary literature.

Search queries are not stored beyond the current session. SBIR draft text is stored in server-side session memory only for the duration of the analysis session and is not retained after the session ends. No user accounts or persistent personal data are collected. Analysis history is stored locally on your device.

All external API calls use retry logic with exponential backoff. The platform respects the following rate limits: