Identifying opportunities for late-stage C-H alkylation with high-throughput experimentation and in silico reaction screening

Goal ord-five-20261013-jordan: internally reviewed submission 5/5; formal human approval remains pending. Submission sub_80ybh0s9lj, verification vrn_1sc50m5j67. Dataset art_cskfjkjam0 SHA256 b57fc0efd0fd361f5c52dda8f6e90bc4f02ddd2182cb2b03f790cfb9a5c2cf28; sources art_l7vtbcfdmd SHA256 3590cf655d44692ffcfb71c89f9635006deb78e4c27f8227a8980d166d00531f. Server hashes match reviewed files.

814 ORD descriptions:621HTE+115new validation+12optimization+18scale-ups with26products+48S7/S9 comparison arms. The48 arms form24pairs and may overlap other experiments; unique physical trial count is unknown. All368decoys+44historical rows excluded explicitly. Missing absolute outcomes are unknown, not fabricated. LCMS fractions are area percentages, uncharacterized regioisomer classes have formula-targeted identifiers rather than invented structures, and scale-up products follow primary drawings with source contradictions attached.

Fresh independent Astra reviewer checked all736SURF joins,2208area values, inputs/conditions,12S8cells,18preps/26drawings and all48arms. Corrected annotation-only comparison coverage, removed unsupported definite vessel assumptions for ambiguous arms, and clarified Indinavir stereo evidence: three depicted source bonds, two centers identified through source name/CAS and primary chemical registries. Reviewer accepted complete available-evidence coverage with no remaining required fixes; actual38e chemical identity and comparative trial deduplication remain source uncertainties. These were preserved, not claimed solved by curation.

Actual ord-schema0.8.3CLI exit0, all37source manifest entries checked, new-directory offline reconstruction byte-identical, required tfp verify preflight passed. MainSI and all reconstruction facts/code/coverage/validation included. Optional NMR original PDF omitted only for5MiB upload limit, with officialURL/hash/downloader retained. Established public organizational curator and AI assistance disclosed, no ORD IDs assigned. No external correspondence or upstream PR.

Next: await the independent human curator stage. Do not count this as formally verified until authoritative verdict. Any actionable feedback requires a continuation attempt, not duplicate pending submissions. Exact final package, reviewer reports, source hashes and all five goal submissions are preserved in the reusable goal checkpoint and delivery bundle.

A fresh independent scientific reviewer has completed QA with no open must-fix findings for the exact frozen dataset SHA-256 b57fc0efd0fd361f5c52dda8f6e90bc4f02ddd2182cb2b03f790cfb9a5c2cf28 and sources ZIP SHA-256 3590cf655d44692ffcfb71c89f9635006deb78e4c27f8227a8980d166d00531f.

Review independently checked 736 physical SURF source joins and all 2,208 analytical areas; all 12 S8 cells; all 26 characterized product structures against primary SI drawings; and all 48 source-described S7/S9 arms, their 24 pair labels and absent absolute outcomes. A clean offline rebuild from the final ZIP reproduced the dataset bytes exactly. Actual ORD0.8.3 CLI validation exited 0. Coverage remains 814 ORD descriptions =766 enumerated/preparative records +48 comparison-arm descriptions, with an unknown number of unique physical trials. The 368 decoys and 44 historical-literature entries remain excluded.

Documented source uncertainties remain visible, notably 38e's conflicting drawing/name/formula/mass evidence and the impossible-to-deduplicate comparison-arm trial identities. These were assessed rather than silently resolved. The Indinavir evidence clarification distinguishes directly drawn stereochemistry from named-drug/CAS-based identity resolution.

This is scientific QA, not formal Tokens for Progress approval. Artifacts are ready for the parent agent's authorized upload/submission workflow; no submission or completed outcome is claimed in this note.

Independent inspection found that SI Figure S2 directly depicts only three of Indinavir's five stereocenters. The source names Indinavir and gives CAS150378-17-9. Three depicted centers were checked (side-chain pentyl C4(R), indanyl C1(S)/C2(R)); piperazine C2(S) and side-chain pentyl C2(S) are not resolved by that drawing. Their assignments follow the source named-drug/CAS identity in primary PubChem CID5362440, corroborated by FDA GSRS UNII9MG78X43ZT and its stereospecific InChIKey CBVCZFGXHXORBI-PXQQMZJSSA-N. The curated SMILES gives that exact key. Sources: https://pubchem.ncbi.nlm.nih.gov/compound/Indinavir and https://precision.fda.gov/uniisearch/srs/unii/9mg78x43zt .

No chemical structure changed. Every affected input and the README now state this distinction explicitly; original non-stereospecific SURF SMILES remains preserved. This evidence clarification changes the exact artifact bytes and supersedes the preceding checkpoint hashes.

Current frozen dataset nippa-2023-minisci.binpb.gz: 158,465 bytes, SHA-256 b57fc0efd0fd361f5c52dda8f6e90bc4f02ddd2182cb2b03f790cfb9a5c2cf28. Current sources nippa-2023-sources.zip: 4,198,739 bytes, SHA-256 3590cf655d44692ffcfb71c89f9635006deb78e4c27f8227a8980d166d00531f. Actual ORD0.8.3 validator again exited 0. Coverage remains 766 enumerated/preparative records plus 48 source-described comparison arms in 24 pairs, with unresolved unique physical count. The fresh reviewer is checking this exact final freeze before any submission.

Fresh independent review identified that the earlier 766-record candidate retained S7/S9 comparison labels only as annotations on HTE records, leaving source-confirmed air and low-concentration arms undiscoverable as reaction descriptions. This revision supersedes that earlier candidate and policy.

The frozen dataset now has 814 ORD records: 766 individually enumerated/preparative records plus 48 source-described comparison arms, linked into all 24 S7/S9 pairs. The source does not allow unique physical trial identities or overlap among these arms, S8 cells and SURF runs to be resolved. Therefore 814 records must not be reported as 814 unique syntheses. Pair labels are preserved verbatim; no absolute outcome measurements or unproved equivalence to existing trials are invented. Conflicting absolute amounts, solvent conditions and scale remain explicit; S8/comparison vessel capacities are left unasserted.

The underlying 766 records remain 621 physical HTE rows, 115 newly conducted literature-validation screens, 12 S8 optimization cells and 18 isolated preparations with 26 characterized products. All 368 computational-decoy and 44 historical-literature rows are excluded with an exact ledger. Audit compares all 736 physical SURF source rows and 2,208 analytical values. Actual ord-schema 0.8.3 CLI validation exited 0 on the revised dataset.

Dataset nippa-2023-minisci.binpb.gz: 157,935 bytes; SHA-256 99b0a07f543e474dd40d4cebf9744edf3abcd781a5a81888ef9fb1878733e9a2. Sources nippa-2023-sources.zip: 4,197,922 bytes; SHA-256 ab9d7e8a9729696100035dc3652e2dd948cb7f60639d968de1ac078280755bdb. The transport archive retains the main SI PDF, every rebuild input, source ledgers, scripts, README and exact validation logs. Only the optional original NMR-spectrum supplement PDF is omitted to fit the 5 MiB transport limit; its exact publisher URL, SHA-256 and reproducible downloader remain included, and its source copy remains in the working checkpoint.

The revised frozen artifacts have been sent for fresh independent validation and clean rebuilding. No upload, submission or completed outcome is claimed at this checkpoint.

Built 766 ordinary reaction records: 621 main HTE +115 new literature-validation HTE +12 SI Figure S8 optimization cells +18 isolated scale-up preparations with26 characterized products. All24 signed S7/S9 paired comparisons are retained as explicitly external source observations, without invented absolute outcomes or trial identity. Excluded368 computational decoys and44 historical literature records; each exclusion is listed.

The compressed dataset SHA-256 is 26096724d76cbc503283d151c261a9ae4dc6bb944e5f39f90f68849a992fb97c. The rebuild sources ZIP SHA-256 is 2d45ca7fe1ce30bbd90798682ed172f59b61b0dc2f23586abf9f84e1c8f9958f. The actual ord-schema0.8.3 CLI validator exited0 on that dataset; stdout/stderr and invocation/hash are archived. An audit compared every one of736 physical SURF records' scales, times, temperatures and2,208 analytical values with the original publisher data, and checked24 comparison observations, exclusions and absent ORD IDs. A separate chemistry agent extracted and checked18 scale-up preparations; full package is now frozen for a fresh independent review. Source ambiguities remain explicit. No upload, submission or upstream ORD PR has yet occurred.

Actual publisher files contain 1,148 total SURF rows: 621 main HTE, 115 new literature-validation experiments, 44 historical literature records, and 368 computational decoys. The 736 physical SURF records have unique source IDs and no overlap between the two physical subsets. These counts differ from prose/card estimates. SI Figure S8 adds 12 quantified optimization records; isolated scale-up extraction finds 18 preparations with 26 named products, despite prose claiming 30 products.

SI Figures S7 and S9 each provide twelve signed comparison labels, not individual absolute reaction outcomes. The figure panels/captions conflict on scale (0.5 versus 5 micromol); S7 also conflicts on concentration and solvent ratio. Preserve all 24 labels and conditions as explicitly external comparison observations indexed by substrate/acid, with unresolved trial identity and no invented absolute outcomes or duplicate physical experiments. Figure S8 has exchanged acid j/o labels relative to structures; retain drawn structures with original labels and conflict notes. Absolute amounts for its 12 records remain unknown because the scale conflict cannot be resolved.

SURF oxidant SMILES has one neutral ammonia beside peroxydisulfuric acid; the named reagent, formula and CAS consistently specify (NH4)2S2O8, so use charge-balanced diammonium peroxydisulfate. Substrate 27's figure and SMILES depict sulfide but its CAS names a sulfoxide; preserve the depicted/source structure and retain contradictory CAS only as metadata. Restore named Indinavir/Avanafil stereochemistry from matching PubChem identities and SI drawings, checking unchanged connectivity. Uncharacterized mono/di products remain formula-defined categories; no regioisomer is guessed.

Confirmed wishlist item126569963 Available. Downloaded the Nature article and all ten linked publisher supplementary files directly with HTTP200, including full experimental SI and seven SURF data/template files. Read pinned curation Skill, card and verifier contract. The paper is CC-BY4.0. Will curate physical experiments conducted for this paper, including HTE, the new literature-validation experiments and isolated scale-up preparations; exclude synthetic decoys and previously published literature reactions. Preserve LC-peak area fractions as semiquantitative analytical outcomes, not isolated yield. Product regioisomers without confirmed connectivity will remain named/formula-defined categories rather than guessed structures. A separate agent is extracting characterized scale-up reactions for integration; a fresh independent reviewer will review frozen artifacts before submission.