Get Started with Tesorai Chat

Tesorai Chat helps you explore proteomics data through conversation. Use it to check data quality, compare experimental groups, create figures and tables, and investigate biological context.

This guide walks through a first analysis. You can start with a completed Search job, your own processed results, or a public study from Trove. Use an analysis template for a guided starting point, or ask your own question.

1. Choose your data

Starting point	How to begin
Completed Tesorai Search job	Open Chat for the job, or use workspace Chat to select the jobs you want to analyze.
Results from another tool	Upload and attach your result files in Chat. See Analyze Your Own Results for supported formats and upload instructions.
Public proteomics study	Find a study in Trove, inspect its details, and select Add to Data or Analyze in Chat. See Using Trove.

Check which jobs or files are selected before asking your question. If you include multiple sources, refer to them by name so Chat knows which to use.

For multiple Search jobs, see Analyze Multiple Search Jobs. If you use Trove, inspect the available sample metadata before deciding which comparisons are possible.

2. Start with a template or your own question

Once you are in Chat, use the available analysis templates and starter cards to help choose a starting point. Templates launch predefined prompts for common analysis tasks, so you do not have to write the initial request from scratch.

Choose a template that fits your data and scientific question. Read its instructions and provide the sample information or other context needed for your experiment. For example, a group comparison needs clear group assignments, while a quality-control analysis starts with checking the selected samples and their measurements.

You can also type your own question directly. After starting with a template, continue the conversation to explore individual findings or request additional figures and tables.

Whichever route you choose, confirm the selected data, sample assignments, and analysis settings before interpreting the results.

3. Describe your experiment and goal

Give Chat the context it needs: what each sample represents, which groups you want to compare, and whether there are technical replicates, fractions, paired samples, or known batches.

For example, adapt this prompt to your own data:

I have three independent control samples and three independent treated samples. Use the attached protein abundance table and sample metadata. Summarize the data and confirm the sample-to-group assignments before running any comparisons. Ask me if an assignment is unclear.

For your own processed results, also explain the source tool and whether abundances have already been normalized or log-transformed. If you are unsure, ask Chat to inspect the file and explain what it can determine.

See the Experimental Design Guide for how Search design fields describe your files.

4. Start with a data-quality question

A useful first step is to understand the data before comparing groups. Ask Chat to summarize sample coverage, missing values, abundance distributions, and relationships between samples.

For example:

Summarize protein coverage and missing values by sample. Plot the abundance distributions and sample correlations. Flag unusual samples and explain what might need investigation before differential analysis.

Review the findings before deciding what to do next. An unusual sample may reflect biology, sample preparation, or measurement quality; investigate the cause before excluding it.

5. Review the analysis approach

For a group comparison, specify the comparison and ask Chat to explain its approach. Where relevant, review filtering, normalization, missing-value handling, the statistical method, and multiple-testing correction.

Compare treated versus control. Before running the analysis, explain your proposed filtering, normalization, missing-value handling, statistical test, and multiple-testing correction. Use independent biological samples as the units of comparison.

Tell Chat about paired measurements or batches explicitly. Technical repeats and fractions should not be counted as additional independent biological samples. If the available metadata does not support the intended comparison, resolve that before proceeding.

6. Explore and refine the results

Once you have reviewed the approach, ask Chat to run the analysis and create the outputs you need. Follow-up questions can focus on a particular protein, change a display setting, or investigate a biological process.

For example:

• “Create a volcano plot and state the significance and fold-change cutoffs.”
• “Show the individual sample abundances for the proteins highlighted.”
• “Which pathways are enriched, and what background set did you use?”
• “Explain which findings come from this dataset and which come from external biological knowledge.”

When an analysis changes, check that the updated figures and tables use the same sample selection and settings.

7. Inspect the code and supporting evidence

Use View code on generated outputs to inspect how the analysis was performed. Check the input data, sample grouping, transformations, statistical settings, and plotting choices.

If a finding depends on a particular Search identification, return to the Search results dashboard to inspect its supporting evidence. The Search Results guide explains the result tables, confidence measures, and intensity columns.

8. Save or export your work

Download the figures and tables you want to keep using the export options on the generated outputs. You can also ask Chat to prepare a results table or summarize the methods used.

For example:

Prepare a downloadable results table and a methods summary listing the input data, sample groups, filtering, normalization, missing-value handling, statistical test, and multiple-testing correction.

Keep the analysis code and relevant settings with your exported results so you and your collaborators can review how they were produced.

Watch the demo

The video below provides an additional introduction to Chat. Interface details may differ from the current platform.

Credits and help

Chat uses the same credit balance as Search and Trove. Usage depends on the complexity of the analysis; see Plans and Credits for details.

For help, email info@tesorai.com with your data type, analysis goal, and a description of where you got stuck.