> ## Documentation Index
> Fetch the complete documentation index at: https://docs.purna.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Bioinformatics
> Turn biological datasets into reproducible analyses, validated results, figures, and downloadable files.
Bioinformatics in Purna combines your biological question, uploaded data, scientific software, and an autonomous analysis workflow in one session. You describe the outcome you need, review the proposed methodology, and receive the scripts, result tables, figures, and written interpretation.
Use it for analyses that require working with the contents of a dataset rather than only searching databases or literature.
## What you can analyze
Depending on your data and the available methods, Purna can help with:
* Dataset inspection and quality control
* Bulk RNA-seq differential expression
* Single-cell and other omics workflows
* Gene, pathway, and enrichment analysis
* Sequence, variant, and genomic interval processing
* Statistical testing and exploratory analysis
* Scientific tables and publication-ready visualizations
The appropriate workflow depends on the structure and quality of your data. Review the proposed design, sample assignments, comparisons, and statistical thresholds before execution.
## 1. Add data and ask a focused question
Attach the dataset to a new or existing session, then describe the analysis you want. Reference the biological groups, comparisons, or outcomes that matter when you know them.
For example:
```text theme={null}
Run differential gene expression analysis on this dataset. Compare senescent
and young samples, account for cell line, use an adjusted p-value threshold of
0.05, and return the complete results, significant genes, QC plots, and scripts.
```
Good requests identify:
* The input files to use
* The experimental groups or conditions
* Relevant covariates, batches, or paired samples
* The required comparison or contrast
* Statistical and fold-change thresholds
* The desired tables, figures, and report format
If important information is missing, Purna can ask follow-up questions before preparing the workflow.
## 2. Review the analysis plan
Multi-step dataset analysis uses [Planning mode](/chat-experiences/planning-mode). Before execution begins, review how Purna proposes to handle:
* Dataset extraction and file identification
* Count matrix, annotation, and sample metadata matching
* Missing values, duplicate identifiers, and data-type checks
* Low-count or quality-based filtering
* Normalization or transformation
* Statistical design, covariates, and contrasts
* Multiple-testing correction and significance thresholds
* Quality-control figures and expected deliverables
Request changes if the sample groups, model design, comparison direction, or validation criteria do not match your experiment. Once approved, the plan is handed to the analysis agent.
## 3. Inspect reproducible analysis files
Purna saves the scripts and outputs created during the analysis. You can open a script to inspect the actual processing steps, model design, contrast, thresholds, and export paths.
In the differential-expression example shown, Purna uses R and DESeq2, applies a low-count filter, includes cell line in the model, and creates both combined and cell-line-specific comparisons.
Common outputs include:
| Output | Purpose |
| -------------------------------- | ------------------------------------------------------------------- |
| Analysis scripts (`.R` or `.py`) | Preserve the executed workflow for review and reuse |
| Complete result tables (`.csv`) | Provide model statistics for every tested feature |
| Filtered gene lists (`.csv`) | Collect genes that meet the approved thresholds |
| QC and result figures (`.png`) | Show sample structure and statistical results visually |
| `report.md` | Document the analysis methods, checks, and findings |
| `answer.md` | Summarize the outcome and direct you to the supporting deliverables |
Open or download these outputs from **Files**. The available file types vary with the analysis.
## 4. Validate sample structure
Quality-control plots help you determine whether the samples behave as expected before interpreting differential results.
The sample-distance heatmap in this example uses variance-stabilized expression values and hierarchical clustering. It helps reveal:
* Whether biological replicates cluster together
* Whether experimental groups separate as expected
* Potential outliers or mislabeled samples
* Cell-line or batch effects that may need to be modeled
Clustering patterns are diagnostic, not proof of a biological effect. Compare them with the experimental design and the other quality-control results.
## 5. Interpret differential-expression results
The volcano plot combines effect size and statistical significance so you can see the overall result and identify genes that meet the approved thresholds.
In the example shown:
* The horizontal axis represents log2 fold change for senescent relative to young samples.
* The vertical axis represents adjusted statistical significance as `-log10(adjusted p-value)`.
* Orange points are higher in senescence.
* Blue points are lower in senescence.
* Grey points do not meet both the approved FDR and fold-change thresholds.
Use the complete result table to inspect the underlying statistics rather than relying on the plot alone.
Depending on the model, result columns can include the feature identifier, normalized mean expression, log2 fold change, standard error, test statistic, p-value, and adjusted p-value. Confirm the comparison direction and filtering rules before using a ranked or significant-gene list.
## 6. Review the answer and continue the analysis
Begin with `answer.md` for the concise interpretation, then open `report.md`, the complete result tables, figures, and scripts to verify how the conclusions were produced.
You can continue with follow-up requests in the same session. For example:
* Change the FDR or fold-change threshold.
* Examine one cell line or subgroup separately.
* Label a different set of genes on the volcano plot.
* Run pathway enrichment on the significant genes.
* Create an additional heatmap or comparison table.
* Export a result in another format.
Purna can reuse the existing plan, scripts, outputs, and session files. If the follow-up materially changes the methodology or requires new capabilities, Purna may prepare another plan for approval.
## Scientific review checklist
Before using the results in a manuscript or decision, confirm that:
* Sample identities and group assignments are correct.
* The model design matches the experiment.
* Important covariates and batch effects are handled appropriately.
* Filtering, normalization, and transformation steps are justified.
* The contrast direction is clearly stated.
* Multiple-testing correction is applied where required.
* Every reported gene satisfies the stated thresholds.
* Key findings are supported by the complete result table.