# Legacy Subnet 25 Mainframe

Mainframe is a decentralised science subnet on Bittensor. It provides computing power and community talent to solve scientific problems.

Subnet 25 currently tackles decentalized protein folding using molecular dynamics (MD) a method for simulating the physical movements of atoms and molecules. 

Protein folding is a complex and computationally expensive problem. Despite its challenges, it holds the key to novel drug discovery and promising medical breakthroughs. Protein engineering is already a $2.6bn market and is set to grow significantly, but the technical barriers remain high, with full simulations often taking days or weeks, making it typically only accessible to the large corporations([Protein engineering market size and share report](https://www.grandviewresearch.com/industry-analysis/protein-engineering-market)).

This combination of computational complexity and cost has attracted the world’s best AI researchers and entrepreneurs. However, while models like AlphaFold offer SOTA protein folding inference, this centralised model entails high costs for computing and for researchers, who are often limited to a fixed number of queries per day. 

Subnet 25 was Bittensor’s first venture into academic use cases, providing a competitive, distributed marketplace for protein folding using molecular dynamics (MD), By offering MD at a more cost-effective rate than centralised rivals, Subnet 25 demonstrates the network’s efficacy and flexibility, proving that we can tackle computationally rich problems with significant market value.

Using [OpenMM](https://openmm.org/), a widely used open-source package for molecular dynamics, Subnet 25 incentivises miners to simulate the folding of proteins into low-energy configurations (analogous to minimizing loss). This aligns with the desired outcome of biologically stable structures, while also being pseudo-deterministic and transparent. The simulation starts with an initial 3D protein structure, places it in a cell-like environment, and models its evolution over time.  When it's run successfully, the simulation can accurately predict the final (or native) 3D structure. This allows researchers to understand the protein’s biological function.

Since launching in June 2024, subnet 25 has already completed over 400,000 protein-folding jobs using GROMACS and [150,000 jobs on OpenMM](https://www.macrocosmos.ai/sn25/dashboard). 

<figure><img src="https://1538249205-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FJDlWdmSC3GnzBPSkAiBM%2Fuploads%2FonQQP7kAjm56Uw5SPUsS%2FScreenshot%202025-03-05%20at%2017.51.22.png?alt=media&#x26;token=89bc209f-e23c-4f9b-bbd9-fd819aba13da" alt=""><figcaption></figcaption></figure>

Our ambition is to produce a decentralized and specialized supercomputer, optimizing the study of proteins and other molecules making it more efficient and accessible, whilst opening us to the scientific research and drug discovery market.

<figure><img src="https://1538249205-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FJDlWdmSC3GnzBPSkAiBM%2Fuploads%2FgzL9mHfqXo4ioaFhWCdQ%2FScreenshot%202025-03-05%20at%2017.52.35.png?alt=media&#x26;token=6c07539d-84c6-4ae5-a459-51468f026a56" alt=""><figcaption></figcaption></figure>

This subnet is designed to empower researchers to conduct world-class drug discovery research utilising deep-learning and molecular dynamics. This allows researchers to:

* Determine *druggability*: Identify pockets suitable for small-molecule binding.
* Optimize compounds: Use MD-guided filters to refine and rank ligands.
* Reduce R\&D costs: Mitigate the need for trial-and-error screening.

<figure><img src="https://1538249205-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FJDlWdmSC3GnzBPSkAiBM%2Fuploads%2F0WMXcsQ1aLzRiLMW81cV%2FScreenshot%202025-03-20%20at%2017.05.12.png?alt=media&#x26;token=0e163ca8-5717-4b1b-8cbd-3049266907c1" alt=""><figcaption></figcaption></figure>

For biotechnology and pharmaceutical companies, MD simulations represent a solid strategy for prioritizing projects and focusing experimental resources where they are most likely to succeed.

Subnet 25 is pioneering cutting-edge science on the blockchain - and exemplifying how Macrocosmos and Bittensor can be deployed to stimulate academically rigorous and commercially significant research.

For more details about the subnet 25's R\&D work, check out our Substack articles:

* [Open-source, autonomous, & collaborative: understanding SN25’s approach](https://macrocosmosai.substack.com/p/open-source-autonomous-and-collaborative)
* [2 months, 70,000 proteins folded: Our roadmap for scaling Subnet 25](https://macrocosmosai.substack.com/p/2-months-70000-proteins-folded-our)
* [Harder, better, faster, stronger: updating our protein folding base miner](https://macrocosmosai.substack.com/p/harder-better-faster-stronger-updating)

Related resources

* [Website](https://www.macrocosmos.ai/sn25)
* [Dashboard](https://www.macrocosmos.ai/sn25/dashboard)
* [GitHub](https://github.com/macrocosm-os/folding)
* [Substack](https://macrocosmosai.substack.com/t/protein-folding)
* [Bittensor Discord](https://discord.com/channels/799672011265015819/1234881153832321024)
* [Macrocosmos Discord](https://discord.com/channels/1238450997848707082)
* [Cosmonauts - Macrocosmos Telegram](https://t.me/macrocosmosai)
* [Macrocosmos X](https://x.com/MacrocosmosAI)


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.macrocosmos.ai/subnets/subnet-25-mainframe.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.