Querying Your Data¶

One of AIVA's most powerful capabilities is querying your uploaded genomic data using natural language. Instead of writing SQL, you describe what you want in plain English, and AIVA translates your request into a database query, executes it, and presents the results.

How It Works¶

When you ask a data-related question, AIVA uses the Genomic Data Query tool behind the scenes:

1. You ask a question in natural language (e.g., "How many variants in my sample have a CADD score above 20?").
2. AIVA generates a SQL query based on your question and the schema of your uploaded sample.
3. The query executes against the database where your parsed data is stored.
4. Results are returned and rendered as a formatted table, a summary, or a chart, depending on what best fits the answer.

Example Queries¶

Here are practical examples grouped by common analysis tasks. Use these as starting points and adapt them to your specific data.

Counting and Summarizing¶

Filtering and Searching¶

Ranking and Prioritization¶

Cross-Sample Analysis¶

Understanding Results¶

AIVA presents query results in the format that best fits the data:

Tables¶

For queries that return rows of data, AIVA renders an interactive table directly in the chat. You can scroll through the results and review individual records.

Summary Statistics¶

For count or aggregation queries, AIVA provides a concise textual summary (e.g., "Your sample contains 42,387 variants, of which 1,204 are classified as pathogenic.").

Charts and Visualizations¶

When a visual representation adds value, AIVA may use the Code Interpreter to generate charts. For example:

• "Plot the distribution of CADD scores" produces a histogram.
• "Show a bar chart of variants per chromosome" produces a bar plot.
• "Create a pie chart of variant consequences" produces a pie chart.

Charts are rendered inline in the conversation and can be viewed in full resolution.

Query Transparency¶

AIVA shows you the tool it used and provides a summary of the operation. This transparency allows you to:

• Verify the query logic: Confirm that the SQL query matches your intent.
• Learn your data schema: See which columns are available and how they are named.
• Refine your questions: If the results are not what you expected, adjust your question using the column names you see in the tool output.

Tips for Effective Queries¶

Be Specific About Your Sample¶

If you have multiple samples, specify which one you are asking about:

• "In the BRCA_panel sample, how many variants are pathogenic?"
• "For my WGS_patient_042 sample, show all stop-gain variants."

Use Genomic Terminology¶

AIVA understands standard genomic terms. Use specific language for more precise results:

• "Missense variants" rather than "point mutations that change the amino acid."
• "gnomAD allele frequency" rather than "how common is this variant in the population."
• "CADD score" rather than "deleteriousness score."

Iterate and Refine¶

Start with a broad query and narrow down:

1. "How many pathogenic variants are in my sample?": Get the overall count.
2. "Which genes are they in?": See the gene distribution.
3. "Show me the ones in BRCA1 and BRCA2 with their ClinVar annotations.": Drill into specific genes.

Combine with Other Tools¶

AIVA can chain multiple tools in a single response. For example:

• "Find all pathogenic BRCA1 variants in my sample and search PubMed for recent publications about them." This uses Genomic Data Query for the data query and Biomedical Literature for the literature search.
• "Calculate the average CADD score per consequence type and plot it as a bar chart." This uses Genomic Data Query for the aggregation and Code Interpreter for the chart.

Limitations¶

Next Steps¶

• Explore all AI tools
• Annotate variants on demand

• Search biomedical literature

• Run custom Python analysis

• See example workflows