How should CRISPRInferredModelEfficacy.csv identifiers be mapped in DepMap Public 26Q1?

Q&A
1

How should CRISPRInferredModelEfficacy.csv identifiers be mapped in DepMap Public 26Q1?

I am working with the DepMap Public 26Q1 files and need to map CRISPRInferredModelEfficacy.csv reproducibly to the screen- and model-level metadata.

The first column in CRISPRInferredModelEfficacy.csv is labelled ModelID, but its 3,180 values appear to be composite screen-like identifiers rather than ordinary DepMap ModelIDs.

Using exact matching only, these identifiers have:

  • 0 matches to the 1,208 ModelIDs in CRISPRGeneEffect.csv;

  • 0 matches to ScreenID in CRISPRScreenMap.csv;

  • 0 matches to SequenceID, ModelID or ModelConditionID in ScreenSequenceMap.csv;

  • 0 matches to the other declared identifiers available in these mapping files.

I have deliberately not used prefix extraction or heuristic parsing because I want the mapping to be reproducible and officially supported.

Could you please clarify:

  1. What entity does the first ModelID column in CRISPRInferredModelEfficacy.csv actually identify?

  2. Is there an official exact-key mapping from these identifiers to CRISPRScreenMap.csv, ScreenSequenceMap.csv or another Public 26Q1 file?

  3. If the identifiers are intended to be parsed, is there an official parsing specification for their components?

  4. Which Public 26Q1 file provides explicit screen-source provenance—for example, Broad versus Project Score—for the rows in ScreenGeneEffect.csv?

  5. Is AchillesScreenQCReport.csv the authoritative source for the CanInclude field used to determine screen eligibility?

The aim is to avoid inferred or heuristic mappings and use only documented Public 26Q1 relationships.

Thank you for your help.

2

Hello KonGur,

Thank you for reaching out.

  1. Each row of the CRISPRInferredModelEfficacy.csv corresponds to a replicate sequence that can be found in the ScreenSequenceMap.csvin the SequenceID column.

2-3: The identifiers in CRISPRInferredModelEfficacy.csv in the ModelID column can be parsed to map to the SequenceID column in the ScreenSequenceMap.csv. Once ModelID is split by “_”, the first two substring indicate the ModelID and pDNABatch of the sequence, which can be found in the ScreenSequenceMap.csv file.The rest or the substring past the second instance of “_” are the SequenceID that can also be found in ScreenSequenceMap.csv.

For example, ACH-000001_KY-1_OVR3_c905R1_KY-1 is an identifier listed in CRISPRInferredModelEfficacy.csv that can be parsed by “_”.

" ACH-000001" is the ModelID and “KY-1” is the pDNABatch. The rest of the string, “OVR3_c905R1_KY-1” would be the SequenceID that can be found in ScreenSequenceMap.csv

  1. The source can be found in ModelCondition.csvin the “DataSource” column that maps the source for each “ModelConditionID”. Each sequence can be mapped to its "ModelConditionID: in the ScreenSequence.csvtable.
  2. “CanInclude” field is determined by “PassesQC” and "“ExcludeFromCRISPRCombined” in ScreenSequenceMap.csv. “CanInclude” is True if a sequence passes QC and “ExcludeFromCRISPRCombined” is set to False. We exclude screens that may not reflect the dependency profile of a given model, such as when a screen was treated with a drug.

Let me know if you have any other questions!

We recognize that CRISPRInferredModelEfficacy.csv can be confusing as each entry corresponds to a sequence, not a model. We will be updating the name of the file and adding details in the README to avoid future confusion in 26Q3.