# Quickstart From zero to your first interactive chart in about two minutes, using **Data Insight**. ## Before you start You're signed in, and you're on the **Data Insight** screen (from the sidebar). You'll see a catalogue of datasets - during beta the public ones you can open are **TCGA** (pan-cancer clinical + genomic) and **CPTAC** (proteogenomic). ## Step 1 - Open a dataset Click **Ask questions about this dataset** on **TCGA**. This opens a fresh chat scoped to that dataset. The empty chat shows a few **starter prompts** - you can click one, or type your own. ## Step 2 - Ask your first question Try this starter prompt: > *"What are the top 10 cancer types in TCGA by case count?"* The assistant writes a query, runs it, and answers with a short summary and a table. You've just explored the dataset without writing any SQL. ## Step 3 - Turn it into a chart Now build on that answer with a follow-up - you don't need to restate the cohort, because the chat keeps the context: > *"Plot that as a bar chart."* You get an interactive bar chart, plus an **auto-generated caption** describing the cohort, counts, and any caveats. Hover for values; the chart is saved and re-renders when you reload. That's the core loop: **ask → read → refine**. Keep going with follow-ups - "only include stage III and IV", "show it as a pie chart instead", "relabel the axis". ## What just happened Behind that simple exchange: 1. The assistant translated your question into an aggregating SQL query against the TCGA dataset. 2. It ran the query and summarised the result. 3. On your follow-up, it mapped the result columns to a chart and rendered it with Plotly - then wrote a caption documenting exactly what was plotted. You never leave the conversation, and every chart traces back to the query behind it. ## A second example Distribution and time-to-event questions work the same way. For instance: > *"Plot Kaplan-Meier overall survival curves for the top 5 TCGA cancer types, and include the log-rank p-value."* returns a survival chart with censoring-aware statistics. ## Next steps - **[Prompting for a plot](/prompting-for-a-plot/)** — the framework for phrasing requests and iterating. - **[Interactive Data Visualisations](/visualizations/)** — all 12 supported chart types.

Quickstart

From zero to your first interactive chart in about two minutes, using Data Insight.

Before you start

You’re signed in, and you’re on the Data Insight screen (from the sidebar). You’ll see a catalogue of datasets - during beta the public ones you can open are TCGA (pan-cancer clinical + genomic) and CPTAC (proteogenomic).

Step 1 - Open a dataset

Click Ask questions about this dataset on TCGA. This opens a fresh chat scoped to that dataset. The empty chat shows a few starter prompts - you can click one, or type your own.

Step 2 - Ask your first question

Try this starter prompt:

“What are the top 10 cancer types in TCGA by case count?”

The assistant writes a query, runs it, and answers with a short summary and a table. You’ve just explored the dataset without writing any SQL.

Step 3 - Turn it into a chart

Now build on that answer with a follow-up - you don’t need to restate the cohort, because the chat keeps the context:

“Plot that as a bar chart.”

You get an interactive bar chart, plus an auto-generated caption describing the cohort, counts, and any caveats. Hover for values; the chart is saved and re-renders when you reload.

That’s the core loop: ask → read → refine. Keep going with follow-ups - “only include stage III and IV”, “show it as a pie chart instead”, “relabel the axis”.

What just happened

Behind that simple exchange:

1. The assistant translated your question into an aggregating SQL query against the TCGA dataset.
2. It ran the query and summarised the result.
3. On your follow-up, it mapped the result columns to a chart and rendered it with Plotly - then wrote a caption documenting exactly what was plotted.

You never leave the conversation, and every chart traces back to the query behind it.

A second example

Distribution and time-to-event questions work the same way. For instance:

“Plot Kaplan-Meier overall survival curves for the top 5 TCGA cancer types, and include the log-rank p-value.”

returns a survival chart with censoring-aware statistics.

Next steps

• Prompting for a plot — the framework for phrasing requests and iterating.
• Interactive Data Visualisations — all 12 supported chart types.