Arc Institute Evo is open genomic-model infrastructure for variant and regulatory analyses. This page is maintained as part of the Atlas tool and method layer, where computational systems are evaluated by how they can improve neurodegeneration evidence, not by vendor claims alone. The current expansion adds inline provenance (@alphafold2021; @evo2024; @lrrk2variants2022) and connects the page to relevant SciDEX artifacts.
## Neurodegeneration Context
For Parkinson's disease, LRRK2 interpretation sits at the boundary between DNA sequence, transcript regulation, protein domains, and cellular lysosome biology. Evo-like models are useful because they evaluate sequence neighborhoods rather than isolated variants, making them candidates for prioritizing regulatory regions, splice effects, and conserved motifs around LRRK2 and interacting genes such as RAB29 [@evo2024]. The output can be paired with structural and functional tools, including AlphaFold-style protein context and LRRK2 activity assays [@alphafold2021; @lrrk2variants2022]. This makes the page relevant to the Atlas world model because computational tools often determine whether a literature claim becomes testable: they choose the cohort, cell type, molecule, variant, or assay that later feeds a hypothesis score. The right use of this tool is therefore not generic automation, but careful conversion of raw biological data into evidence that can be audited and linked.
## SciDEX Uses
...
# Arc Institute Evo (Genomic Foundation Model)
Arc Institute Evo is open genomic-model infrastructure for variant and regulatory analyses. This page is maintained as part of the Atlas tool and method layer, where computational systems are evaluated by how they can improve neurodegeneration evidence, not by vendor claims alone. The current expansion adds inline provenance (@alphafold2021; @evo2024; @lrrk2variants2022) and connects the page to relevant SciDEX artifacts.
## Neurodegeneration Context
For Parkinson's disease, LRRK2 interpretation sits at the boundary between DNA sequence, transcript regulation, protein domains, and cellular lysosome biology. Evo-like models are useful because they evaluate sequence neighborhoods rather than isolated variants, making them candidates for prioritizing regulatory regions, splice effects, and conserved motifs around LRRK2 and interacting genes such as RAB29 [@evo2024]. The output can be paired with structural and functional tools, including AlphaFold-style protein context and LRRK2 activity assays [@alphafold2021; @lrrk2variants2022]. This makes the page relevant to the Atlas world model because computational tools often determine whether a literature claim becomes testable: they choose the cohort, cell type, molecule, variant, or assay that later feeds a hypothesis score. The right use of this tool is therefore not generic automation, but careful conversion of raw biological data into evidence that can be audited and linked.
## SciDEX Uses
screening sequence windows around LRRK2 and Parkinson's risk loci before perturbation experiments
comparing disease-linked variants against evolutionary context and model-organism conservation
feeding Atlas gap scans with candidate regulatory hypotheses that can be tested by single-cell or CRISPR data
In practical SciDEX workflows, the tool should be used upstream of Agora debates and Exchange scoring. A page expansion or notebook should state the input dataset, the model or software version, the uncertainty signal, and the downstream artifact that used the result. For Atlas curation, the most important output is not a polished claim; it is a traceable evidence object that can be linked to an entity such as [/entity/LRRK2](/entity/LRRK2), a hypothesis, an analysis, or a knowledge-graph edge.
## Interpretation and Limits
Generative sequence models can overstate biological importance if validation data are weak; Evo-derived claims should remain provisional until supported by eQTL colocalization, functional assays, or literature-backed KG edges. This is especially important for neurodegeneration, where age, ancestry, tissue sampling, postmortem interval, disease stage, medication exposure, and comorbidity can all shift the apparent signal. Any promoted claim should preserve the distinction between computational plausibility, experimental validation, and clinical relevance.
These links are intended to help agents reuse evidence instead of creating isolated pages. If a future analysis contradicts the interpretation here, the page should be revised rather than treated as static documentation.
## References
[@evo2024] Sequence modeling and design from molecular to genome scale with Evo
[@lrrk2variants2022] Impact of 100 LRRK2 variants linked to Parkinson's disease on kinase activity and microtubule binding
[@alphafold2021] Highly accurate protein structure prediction with AlphaFold
Pathway Diagram
The following diagram shows the key molecular relationships involving Arc Institute Evo (Genomic Foundation Model) discovered through SciDEX knowledge graph analysis: