{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DVLRP5ZARWHODSGCPVWJ2GGYHP","short_pith_number":"pith:DVLRP5ZA","schema_version":"1.0","canonical_sha256":"1d5717f7208d8ee1c8c27d6c9d18d83bd93538042b7f2d81c2a46d6412419fc3","source":{"kind":"arxiv","id":"2605.12784","version":2},"attestation_state":"computed","paper":{"title":"ToolMol: Evolutionary Agentic Framework for Multi-objective Drug Discovery","license":"http://creativecommons.org/licenses/by/4.0/","headline":"ToolMol pairs an LLM agent with RDKit tools inside a genetic algorithm to generate drug ligands that show stronger predicted binding than earlier methods.","cross_cats":["cs.NE","q-bio.QM"],"primary_cat":"cs.LG","authors_text":"Andrew Y. Zhou, Michael K. Gilson, Peter Eckmann, Rose Yu, Sharvaree Vadgama, Sumanth Varambally","submitted_at":"2026-05-12T21:58:14Z","abstract_excerpt":"Advances in large language models (LLMs) have recently opened new and promising avenues for small-molecule drug discovery. Yet existing LLM-based approaches for molecular generation often suffer from high rates of invalid and low-quality ligand candidates, a result of the syntactic limitations of current models with regard to molecular strings. In this paper, we introduce $\\texttt{ToolMol}$, an evolutionary agentic framework for de novo drug design. $\\texttt{ToolMol}$ combines a multi-objective genetic algorithm with an agentic LLM operator that iteratively updates the ligand population. We bu"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.12784","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-12T21:58:14Z","cross_cats_sorted":["cs.NE","q-bio.QM"],"title_canon_sha256":"3b0f22d4f78f908926d79b6342cf94213293ef053f4a62a90d37ed0b98edbd22","abstract_canon_sha256":"183d032fb2a149c2fdec77c716384e863ca0116d474c6b4a16dc8e4e9aacb880"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:09:13.083775Z","signature_b64":"BDfIAiIBDpNd1A4c1clFcWztWovLQMiF6NRIZce/HBYltjF1Svl3L9eHqR67e+6snYMMDpcpGmmCpinsrdAoDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1d5717f7208d8ee1c8c27d6c9d18d83bd93538042b7f2d81c2a46d6412419fc3","last_reissued_at":"2026-05-18T03:09:13.082995Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:09:13.082995Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ToolMol: Evolutionary Agentic Framework for Multi-objective Drug Discovery","license":"http://creativecommons.org/licenses/by/4.0/","headline":"ToolMol pairs an LLM agent with RDKit tools inside a genetic algorithm to generate drug ligands that show stronger predicted binding than earlier methods.","cross_cats":["cs.NE","q-bio.QM"],"primary_cat":"cs.LG","authors_text":"Andrew Y. Zhou, Michael K. Gilson, Peter Eckmann, Rose Yu, Sharvaree Vadgama, Sumanth Varambally","submitted_at":"2026-05-12T21:58:14Z","abstract_excerpt":"Advances in large language models (LLMs) have recently opened new and promising avenues for small-molecule drug discovery. Yet existing LLM-based approaches for molecular generation often suffer from high rates of invalid and low-quality ligand candidates, a result of the syntactic limitations of current models with regard to molecular strings. In this paper, we introduce $\\texttt{ToolMol}$, an evolutionary agentic framework for de novo drug design. $\\texttt{ToolMol}$ combines a multi-objective genetic algorithm with an agentic LLM operator that iteratively updates the ligand population. We bu"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ToolMol achieves state-of-the-art performance on multi-objective property optimization tasks, discovering drug-like and synthesizable ligands that have >10% stronger predicted binding affinity compared to existing methods, evaluated on three protein targets. ToolMol ligands additionally achieve state-of-the-art results in gold-standard Absolute Binding Free Energy scores, gaining over existing methods by over 35%.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that the agentic LLM operator, when equipped with the RDKit toolbox, can consistently execute precise and unbiased ligand modifications that genuinely improve multi-objective fitness without introducing selection biases or invalid structures that undermine the genetic algorithm's search.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ToolMol integrates evolutionary algorithms with agentic LLMs and precise RDKit tools to optimize multi-objective drug properties, yielding ligands with over 10% better predicted binding affinity and 35% gains in absolute binding free energy on three protein targets.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ToolMol pairs an LLM agent with RDKit tools inside a genetic algorithm to generate drug ligands that show stronger predicted binding than earlier methods.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"5d9f3911df1532426f59d0f733b187299167d78b3a9a87964ef161e2e7710673"},"source":{"id":"2605.12784","kind":"arxiv","version":2},"verdict":{"id":"34864f5c-dc1d-457b-b3c8-901bf5748c86","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T04:52:05.971627Z","strongest_claim":"ToolMol achieves state-of-the-art performance on multi-objective property optimization tasks, discovering drug-like and synthesizable ligands that have >10% stronger predicted binding affinity compared to existing methods, evaluated on three protein targets. ToolMol ligands additionally achieve state-of-the-art results in gold-standard Absolute Binding Free Energy scores, gaining over existing methods by over 35%.","one_line_summary":"ToolMol integrates evolutionary algorithms with agentic LLMs and precise RDKit tools to optimize multi-objective drug properties, yielding ligands with over 10% better predicted binding affinity and 35% gains in absolute binding free energy on three protein targets.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that the agentic LLM operator, when equipped with the RDKit toolbox, can consistently execute precise and unbiased ligand modifications that genuinely improve multi-objective fitness without introducing selection biases or invalid structures that undermine the genetic algorithm's search.","pith_extraction_headline":"ToolMol pairs an LLM agent with RDKit tools inside a genetic algorithm to generate drug ligands that show stronger predicted binding than earlier methods."},"references":{"count":46,"sample":[{"doi":"10.1038/nchem.1243","year":2012,"title":"Quantifying the Chemical Beauty of Drugs","work_id":"c299a91c-616f-4a90-8bb6-bf819da740a4","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"A., MacKnight, R., Kline, B., and Gomes, G","work_id":"93614457-269c-442f-aeec-7cd5ac2c4586","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"ChemCrow: Augmenting large-language models with chemistry tools","work_id":"1ceaeb05-7517-4fbd-ba9a-7a63c96b23e6","ref_index":3,"cited_arxiv_id":"2304.05376","is_internal_anchor":true},{"doi":"","year":2026,"title":"El Agente Estructural: An Artificially Intelligent Molecular Editor","work_id":"2564b567-6e74-48cc-83d4-88424f9fa307","ref_index":4,"cited_arxiv_id":"2602.04849","is_internal_anchor":true},{"doi":"10.3389/frhem.2024.1305741","year":2024,"title":"A., Fernandez Prada, D","work_id":"cb7bc3b7-960d-4e36-a6a9-6bf1f128e67d","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":46,"snapshot_sha256":"67773c5f1f41a50b697269d2e41017e12c0a70644ab59dfd09caab470b8393e0","internal_anchors":5},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.12784","created_at":"2026-05-18T03:09:13.083130+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.12784v2","created_at":"2026-05-18T03:09:13.083130+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.12784","created_at":"2026-05-18T03:09:13.083130+00:00"},{"alias_kind":"pith_short_12","alias_value":"DVLRP5ZARWHO","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"DVLRP5ZARWHODSGC","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"DVLRP5ZA","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP","json":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP.json","graph_json":"https://pith.science/api/pith-number/DVLRP5ZARWHODSGCPVWJ2GGYHP/graph.json","events_json":"https://pith.science/api/pith-number/DVLRP5ZARWHODSGCPVWJ2GGYHP/events.json","paper":"https://pith.science/paper/DVLRP5ZA"},"agent_actions":{"view_html":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP","download_json":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP.json","view_paper":"https://pith.science/paper/DVLRP5ZA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.12784&json=true","fetch_graph":"https://pith.science/api/pith-number/DVLRP5ZARWHODSGCPVWJ2GGYHP/graph.json","fetch_events":"https://pith.science/api/pith-number/DVLRP5ZARWHODSGCPVWJ2GGYHP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP/action/storage_attestation","attest_author":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP/action/author_attestation","sign_citation":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP/action/citation_signature","submit_replication":"https://pith.science/pith/DVLRP5ZARWHODSGCPVWJ2GGYHP/action/replication_record"}},"created_at":"2026-05-18T03:09:13.083130+00:00","updated_at":"2026-05-18T03:09:13.083130+00:00"}