Hi @LucaChiesa Just download the NTC release and done some primary data validation, would you please help with the following questions? 1. The dictionary calls SMILES the deduplication key and compound the “BCM compound ID of the first-observed duplicate.” Does each output row represent an original DEL encoding, or should the supplement contain one row per deduplicated canonical SMILES? 2. What exactly does “first-observed duplicate” mean? Were multiple compound IDs or DEL encodings sharing a canonical SMILES reassigned to one representative compound value while their original rows were retained? 3. For repeated compound/SMILES rows, do the rows represent distinct encodings, barcodes, libraries, batches, or replicates? Is a discriminator missing from the published file? 4. How should repeated rows be interpreted when count_NTC or zscore_NTC differs: retain separately, sum counts before normalization, use a representative observation, or apply another BCM aggregation rule? 5. Is the stated count_PGK2 = 0 filter applied to original source rows before representative-ID/SMILES deduplication? Would that explain why 7,403 representative compound values also occur in the original PGK2 export? Also, the description says compounds “bound to PGK2 (count_PGK2 >= 0)”; should that threshold read > 0 or >= 1? 6. Were supplement counts generated from the same processed sequencing output as the original release? If the supplement was reprocessed, were counts changed and were zscore_NTC values recomputed against a different population? 7. The dictionary marks canonical SMILES as required, but 2,300 rows contain empty SMILES. Is this a publication defect, and should those structures be recovered using the representative compound ID or another BCM/library mapping? Cheers, @asfarasimconcerned