Example Workflows¶

This page provides step-by-step workflows demonstrating how to combine AIVA's tools for common genomic analysis tasks. Each workflow shows a sequence of prompts and explains which tools AIVA invokes at each step.

Workflow 1: Variant Interpretation for a Candidate Gene¶

Scenario: You have uploaded a VCF file and want to thoroughly interpret variants in a specific gene.

Steps¶

1. Identify variants in the gene

   "Show me all variants in BRCA1 from my sample, including their consequence and allele frequency."

   Tools used: Genomic Data Query

2. Annotate variants of interest

   "Look up the ClinVar classification and gnomAD frequency for each pathogenic or VUS variant you found."

   Tools used: Variant Annotation

3. Check in silico predictions

   "Get CADD, SIFT, and PolyPhen scores for the VUS variants."

   Tools used: Variant Annotation

4. Search for literature evidence

   "Search for publications about each of these BRCA1 variants."

   Tools used: Biomedical Literature

5. Find clinical trials

   "Are there any recruiting clinical trials for BRCA1-mutated breast cancer?"

   Tools used: Clinical Trials

6. Summarize findings

   "Summarize the evidence for each variant including classification, population frequency, in silico predictions, and literature support."

   Tools used: None (synthesis from previous results)

Workflow 2: Rare Disease Gene Prioritization¶

Scenario: A patient presents with a combination of clinical phenotypes, and you want to identify candidate genes and check your variant data.

Steps¶

1. Map phenotypes to candidate genes

   "A patient presents with microcephaly, seizures, and global developmental delay. What are the top 20 candidate genes?"

   Tools used: Phenotype-Gene Prioritization

2. Search for variants in candidate genes

   "Check my sample for any variants in the top 10 candidate genes from the phenotype-gene prioritization results."

   Tools used: Genomic Data Query

3. Annotate the found variants

   "For any variants found, look up their ClinVar classifications and gnomAD frequencies."

   Tools used: Variant Annotation

4. Review the literature

   "Search for publications linking the genes with variants to microcephaly and seizures."

   Tools used: Biomedical Literature

5. Visualize the results

   "Create a bar chart showing the prioritization scores for the top 10 candidate genes, highlighting which ones had variants in my sample."

   Tools used: Code Interpreter

Workflow 3: Pharmacogenomic Analysis¶

Scenario: You want to identify variants with pharmacogenomic implications and explore drug-gene interactions.

Steps¶

1. Identify pharmacogenes in your data

   "List all variants in known pharmacogenes (CYP2D6, CYP2C19, CYP3A4, DPYD, TPMT, UGT1A1) from my sample."

   Tools used: Genomic Data Query

2. Explore drug-gene interactions

   "For each gene with variants, what drugs are affected? Use the knowledge graph."

   Tools used: Knowledge Graph

3. Check clinical significance

   "Look up the ClinVar and PharmGKB annotations for these pharmacogenomic variants."

   Tools used: Variant Annotation, Web Search

4. Find prescribing guidelines

   "Search for CPIC guidelines related to CYP2D6 and tamoxifen."

   Tools used: Web Search

5. Summarize actionable findings

   "Create a summary table of all actionable pharmacogenomic findings, including the gene, variant, affected drugs, and recommended actions."

   Tools used: Code Interpreter

Workflow 4: Sample Overview and Quality Assessment¶

Scenario: You have just uploaded a new sample and want to understand its contents before detailed analysis.

Steps¶

1. Get basic statistics

   "How many variants are in my sample? Break them down by chromosome, variant type, and consequence."

   Tools used: Genomic Data Query

2. Visualize the distribution

   "Plot the variant count by chromosome as a bar chart, and create a pie chart of variant consequences."

   Tools used: Genomic Data Query, Code Interpreter

3. Assess quality metrics

   "What is the distribution of quality scores? Show me a histogram and the summary statistics."

   Tools used: Genomic Data Query, Code Interpreter

4. Identify high-impact variants

   "How many variants are classified as high impact? List the top 10 by CADD score."

   Tools used: Genomic Data Query

5. Check known pathogenic variants

   "Are there any variants already classified as pathogenic or likely pathogenic in ClinVar?"

   Tools used: Genomic Data Query (if Small Variant Annotation was applied) or Variant Annotation

Workflow 5: Gene Network Exploration¶

Scenario: You found a variant in a gene and want to understand its biological context through interaction networks.

Steps¶

1. Explore the gene's network

   "Show me the protein interaction network for EGFR."

   Tools used: Knowledge Graph

2. Identify drug targets in the network

   "Which proteins in the EGFR network are targetable by approved drugs?"

   Tools used: Knowledge Graph

3. Find pathway context

   "What signaling pathways does EGFR participate in?"

   Tools used: Knowledge Graph

4. Search for variants in network genes

   "Check my sample for variants in any of the genes from the EGFR interaction network."

   Tools used: Genomic Data Query

5. Find supporting literature

   "Search for recent publications about EGFR pathway mutations in lung cancer."

   Tools used: Biomedical Literature, Web Search

6. Find clinical trials

   "What recruiting clinical trials are testing EGFR-targeted therapies?"

   Tools used: Clinical Trials

Workflow 6: Statistical Comparison Across Samples¶

Scenario: You have multiple samples in a project and want to compare variant profiles.

Steps¶

1. Count variants per sample

   "How many variants does each sample in my project have? Show me a comparison table."

   Tools used: Genomic Data Query

2. Compare consequence distributions

   "Compare the distribution of variant consequences across all samples in a stacked bar chart."

   Tools used: Genomic Data Query, Code Interpreter

3. Find shared variants

   "Which variants appear in all samples? List them with their genes and consequences."

   Tools used: Genomic Data Query

4. Statistical comparison

   "Is there a statistically significant difference in the number of missense variants between sample A and sample B? Run a Fisher's exact test."

   Tools used: Genomic Data Query, Code Interpreter

Tips for Building Your Own Workflows¶

• Start broad, then narrow: Begin with overview queries, then drill into specific variants or genes.
• Let AIVA chain tools: You can ask compound questions and AIVA will use multiple tools in sequence.
• Save important findings: Flag variants and add comments as you go.
• Use playbooks: For recurring workflows, create a Playbook that guides AIVA through the steps automatically.
• Export results: Use the export features to save your analysis for reports or downstream use.