This is a short guide to demonstrate enabling Carafe in Skyline and building a Carafe AI-generated and fine-tuned library for a set of target peptides. For this guide we will use a small FASTA of abundant proteins.
Download the attached FASTA file: some-proteins.fasta

Download and extract the following mzML formatted DIA Raw Data and tsv formatted search results from the following link: Carafe Test Data

A special version of Skyline supporting Carafe integration is available here: Carafe Preview Build

After installing the Carafe Preview Build of Skyline, you can build Carafe AI-generated libraries in Skyline using the following steps:

Open Skyline and start with a blank document.

• Click the File menu, then Import > FASTA...

• Click the Settings menu, then click Peptide Settings... and select the Library tab:

• On the Peptide Settings dialog, in the Library tab, click the Build button

The Build Library dialog will appear:

• Give the library a Name and click the Browse... button to choose where to save the library.
• Select Carafe as the Data source.
• Click the Next button

The Build Library dialog will present options for tuning Carafe predictions:

• On this page make sure the option "Build library for:" shows "Current Skyline Document"
• Select "Tuning data source:" as "DIA-NN Report Document"

• Select the DIA Raw Data file you downloaded above as the "MS/MS data"
• Select the report.tsv file you downloaded above as the "DIA-NN report document"

• Click the Finish button

In a new Skyline installation, the first usage of Carafe requires setting up a Python environment:

• Click OK to proceed with the Python installation

To enable python to install and operate properly, the Windows registry key "LongPathsEnabled" must be set on the computer. To change this setting
Administrator privileges are required, or you can change the registry through Windows as an Administrator:

• (See "LongPathsEnabled Wiki Entry" for more information.)

If the computer is equipped with an Nvidia Graphical Processing Unit, a dialog will appear asking if you want to enable Nvidia GPU Computation for significantly faster performance of Carafe modeling. This step requires Administrator privileges to install Nvidia software on the computer.

• If you have Administrator credentials, and want GPU-enabled support click Yes.
• Otherwise, if you don't have Administrator privileges but want to use GPU-enabled processing consult with your System Administrator for help.
• Alternatively, you can click No to use Carafe with CPU processing for building libraries.

If you choose to enable Nvidia GPU Computation you will see another window that will download Nvidia software and guide you through the installation process. Follow the steps presented and please be patient because this can take a long time, but it only has to be done once per computer...

• After Nvidia is downloaded, configured and installed, Skyline will continue with the Python installation.

The Python installation and configuration process can sometimes take 10 to 15 minutes, depending on the speed of the internet connection, please be patient...

If the Python installation completed successfully you should see the following message:

• Click OK.

While building the library, Skyline will present a series of dialog messages, with the last being the following dialog:

Back on Peptide Settings dialog:

• Click the checkbox to enable the newly built library
• Click on Explore...

This brings up the Spectral Library Explorer dialog:

• Click on "B" and "2" in the right vertical toolbar enable B and 2+ fragment ions

• For each peptide ion in the predicted library, Carafe predicted tuned Retention Time (RT) is displayed in the lower right-hand corner, underneath the predicted spectrum.
• You can scroll through the list of peptide ions to visualize the predictions.

Now if you do the "Build Koina Library" steps you can compare the Koina and Carafe built libraries.
Or if you do the "Build AlphapeptDeep Library" steps you can compare the AlphaPeptDeep and Carafe built libraries.

Thank you!