{
  "docId": "019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "docSlug": "8757f1b44ef7f176",
  "documentTitle": "2024 Air Street Capital The State of AI Report 2024",
  "authorId": "AirStreetCapital",
  "authorName": "Air Street Capital",
  "documentKindSlug": "consulting-deck",
  "documentKindLabel": "Consulting deck",
  "sourceTypeSlug": "vc_research",
  "sourceTypeLabel": "VC research",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.777,
  "pageNumber": 51,
  "pageCount": 213,
  "prevPage": 50,
  "nextPage": 52,
  "slideType": "case_study",
  "function": "illustrate_case",
  "density": "dense",
  "nDataPoints": 6,
  "notes": "Includes a process diagram of the model training/inference and bar charts showing editing performance.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "process_diagram",
    "bar_chart_grouped"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0856e1444fb9/51",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9#slide-51",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "A model fine-tuned on Cas9 proteins can generate novel editors that were then validated in human cells.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d14-03fd674d4dc1",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.35,
        "x": 0.62,
        "y": 0.68
      },
      "kind": "chart",
      "text": "Base editing performance at HEK2, T39, and CD3G_1 sites",
      "attrs": null,
      "subkind": "bar-grouped",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "8457cc15-acf3-4094-975b-5e3c40cf948a",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.35,
        "x": 0.62,
        "y": 0.42
      },
      "kind": "diagram",
      "text": "CRISPR-Cas Atlas training and inference process flow",
      "attrs": null,
      "subkind": "process",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "0409affa-9ce5-4578-b080-b477177ca6c0",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.45,
        "w": 0.6,
        "x": 0.03,
        "y": 0.38
      },
      "kind": "list",
      "text": "The CRISPR-Cas Atlas consists of >1M diverse CRISPR-Cas operons, including various effector systems, that were mined from 26.2 terabases of assembled microbial genomes and metagenomes, spanning diverse phyla and biomes.\nGenerated sequences are 4.8x more diverse vs. natural proteins from the CRISPR-Cas atlas. The median identity to the nearest natural protein typically fell between 40-60%.\nA model fine-tuned on Cas9 proteins can generate novel editors that were then validated in human cells. One such editor offered the best editing performance and 71.7% sequence similarity to SpCas9 and was open sourced as OpenCRISPR-1.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "1e00a491-27fd-4c86-bfff-c99e8ccdedff",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "Max A:G editing (%): 71.7%",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d14-07c5e590c0f1",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.6,
        "x": 0.03,
        "y": 0.2
      },
      "kind": "paragraph",
      "text": "We previously profiled how LLMs (e.g. ProGen2) pre-trained on large and diverse datasets of natural protein sequences could be used to design functional proteins with vastly different sequences to their natural peers. Now, Profluent has finetuned ProGen2 on their CRISPR-Cas Atlas to generate functional genome editors with novel sequences that, importantly, were shown to edit the DNA of human cells in vitro for the first time.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f3969615-5ec9-4456-abce-8668454d3985",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.6,
        "x": 0.03,
        "y": 0.13
      },
      "kind": "title",
      "text": "Language models that learn to design human genome editors",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "145f6f6c-b675-4ded-a098-ebe4593467fe",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Chartjunk Elimination",
      "slug": "chartjunk-elimination",
      "agent": "Designer",
      "layer": "slide",
      "matchId": "cc2a5571-6b2f-4d9a-962c-151653f21f07",
      "evidence": "Base editing performance at HEK2, T39, and CD3G_1 sites",
      "confidence": 0.7
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 153,
      "from": 9,
      "beatId": "019dd95a-0682-776c-8e35-5f2398b8d1d0",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Exec summary + Research + Industry sections inventory the year",
      "position": 1,
      "confidence": 78,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 85,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-6d9d8ee06d0a",
      "arcName": "Voyage and Return",
      "arcSlug": "voyage-return",
      "beatName": "The Unknown",
      "beatSlug": "voyage-return-the-unknown",
      "evidence": "Research section explores frontier model uncharted territory",
      "position": 2,
      "confidence": 55,
      "parentBeatName": "Complication",
      "parentBeatSlug": "complication"
    }
  ],
  "loops": [
    {
      "to": 58,
      "from": 45,
      "name": "Zoom In",
      "slug": "06-zoom-in",
      "bestFor": "Technical deep-dives, case studies, detailed analysis",
      "matchId": "019dd95a-07fe-70ce-8d3e-3bc439910222",
      "evidence": "Nobels, AlphaFold3, AlphaProteo, ESM3, gene editors, materials, brain fMRI, speech-from-cortex case studies",
      "position": 7,
      "objective": "Zoom into AI for science domain by domain",
      "structure": "The Big Picture -> Key Area of Focus -> Specific Detail -> Implication",
      "confidence": 80,
      "description": "Start broad, then progressively focus on specific details that prove your point"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}