Agent Mode and Case Mode
Toggle between two distinct modes using the buttons in the top-right corner: Agent Mode is your chat-first interface for exploring general genomics questions, searching databases, and conducting research. You don’t need patient data to use this mode. Your sidebar displays chat history organized by date. Case Mode is your case management interface for working with patient data. Your sidebar shows your cases list where you can create, search, and manage individual cases.Chat
The chat interface is the primary way you interact with Purna. Submit natural language queries and receive conversational responses with citations. All conversations are auto-saved and appear in the sidebar. Start a new chat using the + button.Cases
A case represents a patient or sample in your analysis. Each case contains genetic variants extracted from VCF files and is organized into six tabs:- Overview: Dashboard with charts and key metrics
- Findings: Curated variants of interest from your analysis
- Phenotypes: HPO terms associated with the case
- Variants: Browsable and filterable variant data
- Access: Team permissions and sharing settings
- Settings: Case configuration, files, and audit log
Variants
Variants are individual genetic changes identified during analysis. For each variant, Purna displays:- Gene name and consequence type (missense, frameshift, splice site, intron variant, etc.)
- Genomic position and HGVS notation
- Zygosity (homozygous or heterozygous)
- In silico predictions (SIFT, PolyPhen-2, AlphaMissense)
- Read depth and population allele frequency
- ClinVar classification
- Inheritance pattern and rsID
- Pathogenicity classification (Pathogenic, Likely Pathogenic, VUS, VUS/Conflicting, Likely Benign, Benign)
Variant detail panel
Click any variant to open a detail panel with seven tabs of in-depth information:- Assessment: Pathogenicity predictions from SIFT, PolyPhen-2, and AlphaMissense
- Details: Genomic location, HGVS nomenclature, allele frequencies, and conservation scores
- Research: ClinVar entries, PubMed literature, gene expression, and Phen2Gene data
- ACMG: ACMG/AMP classification criteria with AI-enhanced evidence review
- gnomAD: Population frequency data via the embedded gnomAD browser
- AI Analysis: AI-powered variant interpretation
- Notes: Collaborative notes with visibility control
ACMG classification
Purna supports ACMG/AMP variant classification with 28 standard criteria. The ACMG tab displays active criteria with evidence summaries. Use AI Enhance to automatically evaluate criteria using evidence from ClinVar, PubMed, gnomAD, and functional data. You can also manually reclassify variants and track classification history.Gene panels
Gene panels are virtual collections of curated gene lists used to filter variants. When you create a panel, you specify clinical intent, validation status, inheritance models, and variant types. Apply panels from the variant browser or from chat.Research sources
Control which databases Purna searches using the Sources dropdown at the bottom of the chat. Toggle sources on and off across five categories:Research Papers
Research Papers
PubMed, Nature, Cell, NEJM, Lancet, arXiv, bioRxiv
Clinical Evidence
Clinical Evidence
ClinVar, ClinicalTrials, OMIM, ClinGen, Orphanet
Variant Databases
Variant Databases
gnomAD, GWAS Catalog, dbSNP, VarSome, Franklin
Guidelines
Guidelines
ACMG, NCCN, UpToDate, CDC, WHO
Drugs & Therapies
Drugs & Therapies
PharmGKB, DrugBank, FDA, ClinicalTrials
Findings
Findings are variants of interest marked during analysis. Categorize each finding as:- Primary: Variant directly related to the patient’s presenting phenotype
- Causal: Confirmed to cause the patient’s condition
- Secondary: Related to the case but not the primary explanation
- Incidental: Clinically significant variant unrelated to the case phenotype
Phenotypes
Phenotypes are HPO (Human Phenotype Ontology) terms associated with your case. Add phenotypes through AI Assist (Purna suggests terms based on clinical descriptions) or by manually searching by name or HPO ID. Use phenotypes as Primary or Secondary filters during variant analysis.Clinical reports
Purna generates structured clinical reports from your analysis data. Five report types are available: variant interpretation, clinical findings, case summary, ACMG evidence, and variant comparison. Reports include headings, tables, charts, and callout sections.Artifacts
Artifacts are documents Purna creates during conversations:- Text documents: Clinical summaries, reports, and analysis narratives
- Code files: Python scripts for custom analysis
- Spreadsheets: CSV exports of variant data and findings
AI tools
Purna automatically uses a suite of tools to assist your analysis:| Tool | What it does |
|---|---|
| Variant search | Finds variants in a patient case using your specified filters |
| Variant analytics | Aggregates variant data for statistics and charts |
| Panel search | Looks up gene panels by disease or clinical condition |
| NCBI search | Queries gene, SNP, ClinVar, OMIM, and PubMed databases |
| Chart creation | Generates pie, bar, line, and area charts from your data |
| Code execution | Runs Python code with pandas and matplotlib |
| Clinical reports | Generates structured clinical documentation |
| Web search | Searches the web for additional context |