# Atlas vs BigQuery cohorts

There are different ways to construct a cohort within Sandbox. You can construct your cohort by following steps [here](https://docs.finngen.fi/working-in-the-sandbox/which-tools-are-available/atlas/detailed-guide/how-to-define-a-cohort-in-atlas) wit [Atlas](https://docs.finngen.fi/working-in-the-sandbox/which-tools-are-available/atlas) in Sandbox. You can also construct your cohort using [BigQuery data tables](https://docs.finngen.fi/working-in-the-sandbox/which-tools-are-available/bigquery-relational-database).

For the cohort constructed basically using same criteria in Atlas and BigQuery, you will see that there will be difference in sample numbers due to the following reasons.

**One of the reasons** is that not all medical codes present in [service sector detailed longitudinal](https://docs.finngen.fi/finngen-data-specifics/red-library-data-individual-level-data/what-phenotype-files-are-available-in-sandbox-1/service-sector-data) data are present in the[ medical vocabularies](https://docs.finngen.fi/working-in-the-sandbox/which-tools-are-available/atlas/detailed-guide/code-sets-in-atlas).

Imagine, you are looking for Asthma ICD-8 diagnosis code `49320` present in [service sector BigQuery table](https://docs.finngen.fi/working-in-the-sandbox/which-tools-are-available/bigquery-relational-database). The same ICD-8 diagnosis code is absent in the [medical vocabularies](https://docs.finngen.fi/working-in-the-sandbox/which-tools-are-available/atlas/detailed-guide/code-sets-in-atlas) thereby will not pull any such `FINNGENIDs`  in cohort constructed in Atlas.

**Another reason** is that when a visit (refers to a row in service sector data) is considered the current OMOP conversion takes into account all the medical codes in that visit.

Imagine, you are looking for Asthma ICD-10 diagnosis code `J45.9` present in the [service sector BigQuery table](https://docs.finngen.fi/working-in-the-sandbox/which-tools-are-available/bigquery-relational-database) as `CODE1` and `I10` as `CODE2`. The current OMOP conversion process takes in the combination of both `CODE1` and `CODE2` together thereby making the code `J45.9+I10` which is absent in  [medical vocabularies](https://docs.finngen.fi/working-in-the-sandbox/which-tools-are-available/atlas/detailed-guide/code-sets-in-atlas) and will also be absent in the cohort constructed in Atlas.

**Another reason** is that when important information about which medical vocabulary a code should belong to is missing then OMOP conversion will miss such visits.