# Ecosystem of Ideas — LinkedIn drafts Public-source editorial takeaways, not verbatim quotations or original research results. Drafts only; not posted. [Read the public research ecosystem](https://calyr.ai/ecosystem-of-ideas/) ## Protein function, sparsity and epistasis ### Executive summary Protein-function prediction must handle a vast sequence space with limited experimental observations. Aghazadeh’s primer approaches this problem through Fourier representations, sparsity and epistasis: the effects of sequence changes can depend on their combinations. Sequence variation becomes more useful when interactions remain visible. This is a perspective on representing and sampling function, not evidence that every protein can be predicted from a small dataset. ### Statement for LinkedIn Sequence variation becomes more useful when interactions remain visible. Source authors / speakers: Amirali Aghazadeh; recording with David Brookes. [Primary source](https://www.broadinstitute.org/talks/primer-tbd-4) · [Recording](https://www.youtube.com/watch?v=gxYd1cHmbl8) [Journal context](https://calyr.ai/ecosystem-of-ideas/#protein-function) Connected reading (editorial context): - [Multimodal models of protein function](https://calyr.ai/ecosystem-of-ideas/#multimodal-proteins): Both examine protein function through representations of sequence and related information; they are distinct modelling approaches. - [Protein design using deep learning](https://calyr.ai/ecosystem-of-ideas/#protein-design): Function prediction can inform which designed candidates to examine; prediction and experimental validation remain separate. Editorial interest: CALYR 90% · LITHIOS 10%. For further reading and Research Journal access, contact [hello@calyr.ai](mailto:hello@calyr.ai?subject=Research%20Journal%20reading%20access). ## Protein design using deep learning ### Executive summary Computational protein design connects proposed sequences with structural and functional objectives. Baker’s Broad/MIT colloquium presents this research direction through deep learning. Design creates candidates. Measurement tests their properties. The presentation is a public research reference; this summary does not establish independent experimental validation of any CALYR design. ### Statement for LinkedIn Design creates candidates. Measurement tests their properties. Source authors / speakers: David Baker. [Primary source](https://www.broadinstitute.org/videos/protein-design-using-deep-learning-david-baker) · [Recording](https://www.youtube.com/watch?v=-H27Kv5duYA) [Journal context](https://calyr.ai/ecosystem-of-ideas/#protein-design) Connected reading (editorial context): - [Protein function, sparsity and epistasis](https://calyr.ai/ecosystem-of-ideas/#protein-function): Function prediction can inform which designed candidates to examine; prediction and experimental validation remain separate. - [Multimodal models of protein function](https://calyr.ai/ecosystem-of-ideas/#multimodal-proteins): Structure and function representations connect these topics; generating a candidate and predicting its properties are different tasks. Editorial interest: CALYR 80% · LITHIOS 20%. For further reading and Research Journal access, contact [hello@calyr.ai](mailto:hello@calyr.ai?subject=Research%20Journal%20reading%20access). ## Scientific agents for chemistry ### Executive summary Scientific agents extend language models by connecting them to specialist tools. The Cox and White session discusses literature work, chemistry and molecular-dynamics workflows, including ChemCrow and MDCrow. A scientific agent is only as useful as the tools and checks around it. A tool-using workflow still requires task-specific evaluation; successful execution alone does not establish a correct scientific conclusion. ### Statement for LinkedIn A scientific agent is only as useful as the tools and checks around it. Source authors / speakers: Sam Cox; primer by Andrew White. [Primary source](https://www.broadinstitute.org/videos/sam-cox-automating-chemistry-llm-based-agents-primer-andrew-white) · [Recording](https://www.youtube.com/watch?v=XzpJaNhEKyw) [Journal context](https://calyr.ai/ecosystem-of-ideas/#chemistry-agents) Connected reading (editorial context): - [Biomedical research agents](https://calyr.ai/ecosystem-of-ideas/#biomedical-agents): Both discuss language-model agents using scientific tools; domain-specific tasks and evaluations remain distinct. - [Biomni: public project reference](https://calyr.ai/ecosystem-of-ideas/#biomni): Both connect language-model reasoning to specialist tools and execution; their tool environments and scientific scope differ. - [Text2BIM: language-assisted building models](https://calyr.ai/ecosystem-of-ideas/#text2bim): Shared workflow: language-model agents call domain tools or APIs and iteratively check outputs. Chemical and building-model validity do not transfer between domains. Editorial interest: CALYR 80% · LITHIOS 20%. For further reading and Research Journal access, contact [hello@calyr.ai](mailto:hello@calyr.ai?subject=Research%20Journal%20reading%20access). ## Multimodal models of protein function ### Executive summary Protein-function modelling can combine more than amino-acid sequences alone. The Zitnik, Queen and Huang presentation describes a multimodal approach connecting sequence and structure models with textual protein descriptions. Sequence, structure and descriptions offer complementary views of protein function. Such representations support research questions; a model response remains distinct from an independently measured property. ### Statement for LinkedIn Sequence, structure and descriptions offer complementary views of protein function. Source authors / speakers: Marinka Zitnik, Owen Queen and Yepeng Huang. [Primary source](https://www.broadinstitute.org/videos/multimodal-protein-language-models-deciphering-protein-function) · [Recording](https://www.youtube.com/watch?v=LcLmvtXHI1s) [Journal context](https://calyr.ai/ecosystem-of-ideas/#multimodal-proteins) Connected reading (editorial context): - [Protein function, sparsity and epistasis](https://calyr.ai/ecosystem-of-ideas/#protein-function): Both examine protein function through representations of sequence and related information; they are distinct modelling approaches. - [Protein design using deep learning](https://calyr.ai/ecosystem-of-ideas/#protein-design): Structure and function representations connect these topics; generating a candidate and predicting its properties are different tasks. Editorial interest: CALYR 90% · LITHIOS 10%. For further reading and Research Journal access, contact [hello@calyr.ai](mailto:hello@calyr.ai?subject=Research%20Journal%20reading%20access). ## Biomedical research agents ### Executive summary Biomedical research often spans literature, databases, analysis tools and experimental protocols. Huang’s Biomni presentation describes an agent architecture that combines planning, retrieval and executable workflows; Wang’s primer discusses approaches to biomedical discovery agents. Connect tools into a workflow, and keep the evidence traceable. Reported demonstrations belong to the cited work and are not independent CALYR benchmark results. ### Statement for LinkedIn Connect tools into a workflow, and keep the evidence traceable. Source authors / speakers: Kexin Huang; primer by Hanchen Wang. [Primary source](https://www.broadinstitute.org/videos/mia-kexin-huang-general-purpose-biomedical-ai-agent-primer-hanchen-wang) · [Recording](https://www.youtube.com/watch?v=nV7Ltp-M5jM) [Journal context](https://calyr.ai/ecosystem-of-ideas/#biomedical-agents) Connected reading (editorial context): - [Biomni: public project reference](https://calyr.ai/ecosystem-of-ideas/#biomni): The talk and repository describe the same Biomni project. They provide complementary context, not independent validation. - [Scientific agents for chemistry](https://calyr.ai/ecosystem-of-ideas/#chemistry-agents): Both discuss language-model agents using scientific tools; domain-specific tasks and evaluations remain distinct. Editorial interest: CALYR 90% · LITHIOS 10%. For further reading and Research Journal access, contact [hello@calyr.ai](mailto:hello@calyr.ai?subject=Research%20Journal%20reading%20access). ## Biomni: public project reference ### Executive summary The public Biomni project provides a concrete reference for connecting biomedical reasoning with retrieval, planning and code execution. It offers a way to examine how a research question becomes a sequence of tool-assisted operations. Executable research needs sources, tools and a reproducible record. The public repository documents the project; this entry does not claim that CALYR has reproduced its benchmarks or validated it for a specific experimental task. ### Statement for LinkedIn Executable research needs sources, tools and a reproducible record. Source authors / speakers: Stanford SNAP / Biomni contributors. [Primary source](https://github.com/snap-stanford/Biomni) [Journal context](https://calyr.ai/ecosystem-of-ideas/#biomni) Connected reading (editorial context): - [Biomedical research agents](https://calyr.ai/ecosystem-of-ideas/#biomedical-agents): The talk and repository describe the same Biomni project. They provide complementary context, not independent validation. - [Scientific agents for chemistry](https://calyr.ai/ecosystem-of-ideas/#chemistry-agents): Both connect language-model reasoning to specialist tools and execution; their tool environments and scientific scope differ. Editorial interest: CALYR 80% · LITHIOS 20%. For further reading and Research Journal access, contact [hello@calyr.ai](mailto:hello@calyr.ai?subject=Research%20Journal%20reading%20access). ## Text2BIM: language-assisted building models ### Executive summary Text2BIM studies how language-model agents can generate editable building models through specialist software. It connects an expressed design intention to model construction and iterative checking. An editable model is a starting point for checking a design. The reported research demonstrations do not establish general structural safety, building-physics validity or regulatory compliance. ### Statement for LinkedIn An editable model is a starting point for checking a design. Source authors / speakers: Changyu Du, Sebastian Esser, Stavros Nousias and André Borrmann. [Primary source](https://arxiv.org/abs/2408.08054) [Journal context](https://calyr.ai/ecosystem-of-ideas/#text2bim) Connected reading (editorial context): - [Scientific agents for chemistry](https://calyr.ai/ecosystem-of-ideas/#chemistry-agents): Shared workflow: language-model agents call domain tools or APIs and iteratively check outputs. Chemical and building-model validity do not transfer between domains. Editorial interest: CALYR 30% · LITHIOS 70%. For further reading and Research Journal access, contact [hello@calyr.ai](mailto:hello@calyr.ai?subject=Research%20Journal%20reading%20access).