Back to projects
Fullstack

Novagenic — Clinical pharmacogenomics

Screenshot of the Novagenic — Clinical pharmacogenomics project

About the project

Novagenic Clínica — From the lab's spreadsheet to the patient's report

The lab delivers the chip run as a spreadsheet: dozens of samples across hundreds of SNPs, with no names and no interpretation. Novagenic Clínica turns that into a report a physician can read: it calls each gene's diplotypes, looks up the phenotype and recommendations from the CPIC guidelines, and adjusts the result for the medication the patient is already taking. The redesign began with a full technical audit of the previous platform and a working 20-screen prototype.

How it works

Physician

  • Uploads the run's spreadsheet; the parser detects sample columns, reads genotypes by rsID, and validates the structure
  • Assigns each sample to a patient on a mandatory mapping screen: no confirmed link, no report
  • Records and confirms current medication, with brand-name recognition for the Mexican market
  • Generates the report, reviews it on screen, and downloads a PDF with its disclaimers and the version of the tables used
  • Prescription lookup: asks about a drug the patient is not yet taking and sees the result before prescribing it

Patient and administrator

  • The patient signs in with their own account and sees only their report and PDF
  • The administrator manages physicians, patients, and admins, with search, pagination, and export
  • Bulk reassignment of a physician's patients and deletion of lab runs
  • Role isolation verified: a patient cannot read others' reports and a physician cannot see another's patients

The scientific engine

  • CPIC/ClinPGx knowledge base precompiled into versioned JSON: 68 SNPs, 37 pharmacogenes, and their recommendations
  • Allele calling with CPIC definitions scoped to the panel, including a complementary-strand attempt with its own flag
  • Phenoconversion: co-medication cross-referenced against the FDA inhibitor and inducer table adjusts the phenotype
  • Catalog of 239 drugs and 587 recognized names, Mexican brand names included
  • Limitations printed in the report itself: targeted panel without CNV, strand ambiguity, and key-SNP calling