Canonical: https://instain.ru/en/ai/esm3/
Language: en

Biohub, formerly EvolutionaryScale  Language model for protein research

# ESM3. 

Biohub model for protein sequence, structure, and function

4.15 / 5    [Directory score](https://instain.ru/en/ai/esm3/#editorial-rating) [Official website ](https://biohub.org/blog/esm3-simulating-evolution/)

ESM3 studies the relationship between sequence representations, structure, and function. The model is used for computing protein data.

![Biohub publication with ESM3 model presentation](https://instain.ru/assets/images/esm3-screen-36091bdf48ab-1200.webp)

Biohub publication with the ESM3 model presentation.  [Open the source ](https://biohub.org/blog/esm3-simulating-evolution/)

## What the model works with

For educational acquaintance with ESM3, you can take an open protein sequence with a published annotation. The model works with representations of sequence, structure and function. This allows us to consider the connections between them computationally. Before starting, specify which fields the selected SDK option supports. For verification, a small example is convenient in which the original sequence is known and the result can be compared with it by position.

## Check on a known example

Mark a few positions and compare the predicted output with the original sequence. Record which information remained visible in the input. If you add structure data, keep that run separate from the sequence-only run. Differences can show how the input changes the computational prediction and help verify that the library interpreted the format correctly.

## How to read the result

Save the release ID, parameters and origin of the data. Verify the numbering of positions so that missing a character or other version of the sequence does not look like a model error. If an assessment is needed for several examples, choose entries with different features and a known annotation. The model output should be described as computational observation on this data. Protein properties and applicability to real-world studies require independent sources and separate verification. Describe the matches along with the initial conditions so that another reader can repeat the comparison.

## Suitable tasks

- Examine the recovery of hidden positions in public sequence

- Compare model output with annotated data

Features 

## Strengths

- The model takes into account several representations of protein

- The developer publishes the SDK and a description of the operation mode

Things to consider 

## Limitations

- A prediction does not establish a protein’s function or properties

- Weight and access terms depend on the version you choose

## Selection and conditions

Free access

Not confirmed

Trial period

Not confirmed

No sign-up required

Not confirmed

Russian interface

Not confirmed

Local installation

Yes

API

Yes

“Not confirmed” means there is not enough information. The feature may still be available; check the developer’s terms.

Input formats

Data

Result

Data

### Terms

Local weights and hosted access have different terms. Check the release documentation for usage restrictions. A prediction does not establish a protein’s function or properties.

From task to first result

## Practical example

Task

Examine the recovery of hidden positions in public sequence

What to prepare

Public training protein sequence with known annotation and marked positions.

### Task brief or prompt

Computationally compare the predicted tokens in hidden positions with an open source. Save the model parameters.

### Expected result

Table of observations for reproducible educational analysis.

### How to check the result

- Verify the version and sequence numbering.

- Do not accept a computational match as a confirmation of the property.

An editorial teaching example of a task brief, not the result of hands-on testing of the service. Check that the required mode is available, that you have rights to the source material and that the intended use of the result is permitted.

Check details

## Official sources

-  [ESM3 Developer Description ](https://biohub.org/blog/esm3-simulating-evolution/)

-  [Biohub ESM ](https://github.com/Biohub/esm)

-  [Official Biohub portfolio and access ](https://biohub.org/ai-models/)

This description is based on the developer’s material. Features, prices and availability may change; check the primary source before use.

About the tool

Developer

Biohub, formerly EvolutionaryScale

Category

[Code](https://instain.ru/en/category/code/)

Access

Biohub documents hosted access and provides weights for selected local variants. Check model availability and licensing for your chosen setup.

Payment

Local weights and hosted access have different terms. Check the release documentation for usage restrictions.

API

Yes, from the developer

[Developer terms ](https://biohub.org/ai-models/)

Directory score

4.15 / 5 

Features

4.50 / 5 

Workflows

3.60 / 5 

Ecosystem

4.20 / 5 

Documentation

4.30 / 5 

Scores are an editorial guide to the directory. They are not user reviews or the results of personal testing.

[See the methodology ](https://instain.ru/en/methodology/)