Balto

Quick start

Balto is an AI assistant for molecular modeling. Balto speeds up drug discovery by providing users easy access to cutting-edge modeling tools, such as pocket finding and protein-ligand docking. In particular, Balto removes the need for investigators to learn complex tools and helps researchers move from hunches to hits in minutes.

We are constantly expanding and refining Balto's capabilities. Please contact support to share feature requests and suggestions.

UI Organization

After logging into Balto, you will see a window that looks like below, organized into three vertical panels.

• The left panel shows your recent sessions with Balto. You can collapse the left panel by clicking the button at the top-right of the middle panel.
• The middle panel shows your current conversation with Balto. When you first log in, this panel shows several examples of prompts you can ask Balto.
• The right panel shows your workspace. Your workspace shows Balto’s outputs, and the steps Balto took to answer your queries.

Getting help and feedback

You can get help and provide feedback on Balto in multiple ways.

Click the ? button

Pressing the ? button at the bottom-right of the screen opens a menu that contains links to frequently asked questions, an option to book a call with our customer support team, and a link to our Support center.

Feedback via the up/down vote button

To help us improve Balto, please use the buttons below each of Balto’s responses to indicate whether the response was helpful or not.

Tasks that Balto can help you with

Balto is equipped with a suite of tools for several tasks in chemistry, biology, and drug discovery. Here are some of the tasks Balto can assist you with:

Drug discovery

• Query chemical databases
  • ChEMBL database: Query the ChEMBL database for molecules, targets, assays, and activities. Fetch details such as molecular weights and biological activities.
• Similarity between molecules
  • Molecular similarity: Calculate the Tanimoto coefficient between molecules to characterize their structural similarity.
• Patents
  • Check whether a molecule has been patented: Check if a compound is patented based on its SMILES string.
• Calculate protonation sates and stereoisomers
  • Protonation states: Predict protonation states of molecules at specific pH levels.
  • Isomer generation: Generate stereoisomers of molecules.
• Predict properties of molecules
  • LogP, LogD, LogS: Predict the log partition coefficient (LogP), distribution coefficient (LogD), and aqueous solubility (LogS) for compounds.
  • Functional groups: Identify the functional groups within a molecule.
  • Molecular weight: Calculate the molecular weight of a compound.
  • QED: Estimate the Quantitative Estimate of Drug-likeness (QED) of a compound.
• Predict toxicity of molecules
  • CYP interactions: Predict binding interactions of compounds with cytochrome P450 enzymes, which play a key role in metabolizing drugs.
  • hERG blocking: Assess the likelihood of a compound blocking the hERG channel, which is important for cardiac safety.
  • Ames test: Predict the mutagenicity of compounds using the Ames test.
• Plan the synthesis of molecules
  • Synthesis pathways: Generate likely pathways for synthesizing target molecules from precursor compounds.
  • Synthetic accessibility: Calculate the Synthetic Accessibility Score (SAS) to determine how easily a molecule can likely be synthesized.

Biology

• Retrieve, predict, and visualize protein structures
  • Retrieve experimentally-determined protein structures: Search for experimentally-determined protein structures from the Protein Data Bank (PDB) using their UniProt accession codes.
  • Retrieve computationally-predicted protein structures: Load the predicted conformations of proteins from the AlphaFold database using their UniProt accession codes.
  • Visualize proteins: Display the structures of proteins in 3D.
• Identify binding pockets and dock compounds
  • Find pockets: Identify binding pockets within protein structures.
  • Docking compounds: Simulate the docking of small molecules into protein pockets, providing binding poses, energies and visualizations.
• Other
  • Perform various operations on protein structures, such as extracting chains, deleting residues, and converting files.

Academic and research support

• Summarize journal articles
  • Summarize scientific papers by extracting key methods, results, and conclusions.
  • Provide critical analysis of research papers, highlighting strengths and weaknesses.
• Visualize chemicals, proteins, and docking poses
  • Visualize the structures of chemicals in 2D.
  • Visualize the structures of proteins and predicted docking poses in 3D.

Other

• Retrieve the CAS IDs of compounds
  • Retrieve Chemical Abstracts Service (CAS) identifiers for compounds from their names.
• Search the web
  • Perform web searches to retrieve real-time information, current events, and newly published articles related to chemistry, biology, and drug discovery.

Example workflow: Dock a ligand to a protein

In this example we will dock a known ligand (identified by its SMILES string) to a protein (identified by its PDB ID).

In this example, we will explore tyrosine-protein kinase JAK2 (Janus kinase 2). JAK2 is a non-receptor protein tyrosine kinase that plays a key role in cell growth and development.

First, ask Balto to find pockets in its structure by entering the following prompt:

Balto will then respond with the following (your exact response may vary):

To find the pockets for the 3E64 protein, I'll follow the steps below:

1. Download the 3E64 protein file from the Protein Data Bank (PDB).
2. Verify the structural quality and provide details such as resolution and method.
3. Use the downloaded protein file to identify pockets.
4. Visualize the identified pockets in the protein structure.

First, I'll download the 3E64 protein file. Let's begin.

Step 1: Download the 3E64 Protein File

Executing the plan to download the 3E64 protein file will provide the path for pocket finding.

I'll start by estimating the function calls and cost required for this operation.

Identified Pockets for the 3E64 Protein

Here are the details of the pockets identified in the 3E64 protein, including their drugability scores and volumes:

Pocket Name	Drugability Score	Volume (ų)	Hydrophobicity	Polarity	Total SASA (Ų)	Apolar SASA (Ų)	Polar SASA (Ų)	Polar/Apolar SASA Ratio
Red	0.87	1489.0	10.66	40.0	3779.09	2067.52	1711.56	0.83
Green	0.66	175.0	32.0	10.0	935.80	440.73	495.08	1.12
Blue	0.13	87.0	33.18	7.0	545.83	265.85	279.98	1.05
Yellow	0.12	106.0	23.25	8.0	728.11	357.82	370.30	1.03
Magenta	0.06	207.0	11.24	11.0	829.28	413.73	415.56	1.00

Note:

• Drugability Score: A higher score indicates that the pocket is more likely to bind drug-like molecules effectively.
• Volume: Measures the space (in ų) within the pocket, which can correlate with its capacity to accommodate ligands.
• Hydrophobicity and Polarity: These properties impact the binding affinity and specificity of ligands.

If you wish to dock any ligands into these pockets or need further analysis, please let me know!

Once Balto has executed the tasks outlined above, Balto will then show the information it retrieved from the Protein Data Bank:

Balto’s output will include a visual representation of the pockets it found in the protein:

Next, ask Balto to dock **Baricitinib (SMILES: **CCS(=O)(=O)N1CC(CC#N)(n2cc(-c3ncnc4[nH]ccc34)cn2)C1) to the red pocket by. Balto will then respond with the following:

To dock the ligand CCS(=O)(=O)N1CC(CC#N)(n2cc(-c3ncnc4[nH]ccc34)cn2)C1 to the red pocket of the 3E64 protein, I'll follow these steps:

1. Dock the ligand into the red pocket using the specified SMILES string and the red pocket's properties.
2. Visualize the docking result.

Let's begin by docking the ligand 'CCS(=O)(=O)N1CC(CC#N)(n2cc(-c3ncnc4[nH]ccc34)cn2)C1' into the red pocket.

Step 1: Dock the Ligand into the Red Pocket

Proceeding with the docking process.

Balto’s response will include a visualization of the ligand in the optimal predicted pose within the red pocket:

After docking is complete, Balto will share a report of the results:

Docking Results

The ligand 'CCS(=O)(=O)N1CC(CC#N)(n2cc(-c3ncnc4[nH]ccc34)cn2)C1' was successfully docked into the red pocket of the 3E64 protein. Here are the details:

• Binding Energy: -7.17 (moderate binder)
• Visualization: The docked ligand and protein structure have been visualized on the right-hand side of the user's screen.

Docked Ligand File Location: storage://srinivas_deeporigin_com_66ec1b7bff62ef0f0ff7f409/proteins/3E64/docking/red/49739/3E64_docking_top_solution.mol2

If you need further analysis or additional docking, please let me know!

This example workflow shows how Balto can be used to perform complex tasks without writing a single line of code. To explore additional tasks and workflows, simply ask Balto for help.

Happy drug hunting!