A single-cell atlas characterizes dysregulation of the bone marrow immune microenvironment associated with outcomes in multiple myeloma

Description

Technical info

Two .RDS files are included in this version of the release.

• Discovery object: DISCOVERY_ONLY_MMRF_IA_zenodo_public_simplified.rds contains all aliquots belonging to the 'discovery' cohort as described in the paper. This represents the dataset used for initial clustering, cell annotation, and analysis. If using this dataset to transfer cell type labels, we recommending using this version of the object.
• Discovery + Validation object: COMBINED_DISC_VAL_MMRF_IA_zenodo_public_simplified.rds contains both aliquots belonging to the initial 'discovery' cohort, and aliquots belonging to the 'validation' cohort. The group each cell is derived from is listed under the 'cohort' variable. Labels related to cell annotation, including doublet status, are derived from a label transfer process as described in the paper. Labels for the original 'discovery' cohort are unchanged. UMAPs have been reconstructed with both the discovery and validation cohorts integrated. This version of the object was used as the basis for follow-up studies, including the longitudinal analysis. 

--

The discovery object contains two assays:

• "RNA" - The raw count matrix
• "RNA_Batch_Corrected" - Counts adjusted for the combination of 'Study_Site' and 'Batch'.
• Analysis should prefer the original RNA assay, unless using pipelines which does not support adjusting for technical covariates.

The validation object only includes the uncorrected RNA assay.

--

The object contains two umaps in the reduction slot:

• umap - will render the UMAP for the full object with all cells.
• umap.sub -contains the UMAP embeddings for individual 'compartments', as indicated by ‘major_cell_types’ column

--

Outside of the additional batch corrected assay, both versions of the object contain the same set of metadata. The primary difference will be that cell type labels in the validation cohort are derived via label transfer

• sample_id
  • <d_visit_specimen_id>_<aliquot_id>
  • Example: MMRF_1234_2_MMRFA-012345
  • MMRF public_id: MMRF_1234
  • MMRF d_visit_specimen_id: MMRF_1234_2
  • MMRF aliquot_id: MMRFA-012345
• Combines the MMRF visit identifier and the aliquot identifier
• This is unique to each individual bone marrow single-cell RNA aliquot (aliquot_id) processed. A single visit (d_visit_speciment_id) may generate multiple aliquots in the event of technical replicates. A single patient (public_id) may be associated with multiple visits each of which obtained their own sample.

For additional information on standard MMRF sample identifiers:

• public_id
• Indicates a specific patient (n=263).
• MMRF_####
• This is a standard identifier which is used across all MMRF CoMMpass datasets
• public_ids can map to multiple d_visit_specimen_ids and aliquot_ids
• d_visit_specimen_id
  • Y is a number indicate that this is the 'Y' sample obtained from said patient. This does not correspond to a specific timepoint.
• Indicates a specific visit by a patient (n=358)
• MMRF_####_Y
• This is a standard identifier, which is used across all MMRF CoMMpass datasets
• d_visit_specimen_id uniquely maps to one public_id
• d_visit_specimen_id can map to multiple aliquot_ids
• In this cohort, a sample with a visit number of ‘1’ (d_visit_specimen_id ending in 1) will correspond to a sample obtained prior to therapy, where as other values will correspond to a follow-up sample. Please be aware that this is not a general rule for all CoMMpass samples, and to refer to the patient visit table in the MMRF IA clinical flat files for more precise information.
• aliquot_id
• Refers to the specific bone marrow aliquot sample processed (n=361)
• MMRFA-######
• This is a unique identifier for each processed scRNA-seq sample. 
• As of now, is an identifier specific to the MMRF ImmuneAtlas

Each cell has the following annotation information:

• minor_cell_types
  • 'NkT.2.2', indicates this cell is in the 'Natural Killer + T Cell compartment', was originally part of 'Cluster 2', and then was further separated into a refined subcluster 2.2'
  • If a parent cluster did not need to be further seprated, the 'Compartment-subcluster' part is omitted (e.g., 'NkT.6')
  • For most immune cell types, clustering was based on embeddings corrected for 'siteXbatch'. For Plasma, clustering was performed on embeddings corrected on a per-sample basis.
• These refer to an individual cluster derived from 'Seurat'.
• The names in here match those listed in the Cell Annotation Dictionary.
• Format is 'Compartment'.'Compartment-cluster'.'Compartment-subcluster'
• As of now, this uniquely maps to a specific cellID_short annotation.
• Clustering was done on a per compartment basis
• major_cell_types
  • NkT (T cell + Natural Killer Cells)
  • Myeloid (Monocytes, Macrophages, Dendritic cells, Neutrophil/Granulocyte populations)
  • BEry (B Cell, Erythroblasts, bone marrow progenitor populations, pDCs)
  • Ery (Erythrocyte population)
  • Plasma (Plasma cell populations)
• These refer to one of five major compartments as identified roughly on the original UMAP. Clustering was performed on a per-compartment basis following a first pass rough annotation.
• The possible compartments are
• One cluster was isolated from all other populations, and was not assigned to a compartment. This cluster is labeled as 'Full.23', and has the major_cell_type of “Full” (Cluster derived from clustering across all major cell types)
• Each compartment has it's own UMAP generated, which can be accessed in the 'umap.sub' reduction slot
• cellID_short
• This is the individual annotation for each cluster.
• Please see the 'Cell Population Annotation Dictionary' for further details.
• If different seurat clusters were assigned similar annotations, the celltype annotation will be appended with a distinct cluster gene, or with '_b', '_c'
• isDoublet
• This is a binary 'True' or 'False' derived from manual review of clusters following doublet analysis, as described in the paper.
• True indicates the cluster was determined to be a doublet population.
• This is derived from 'doublet_pred', in which 'dblet_cluster' and 'poss_dblet_cluster' were flagged as doublet populations for subsequent analysis.
• In the validation object, the doublet status of new samples were inferred by if label transfer from the discovery cohort mapped the cell from the new sample as one of the previously identified doublet populations. The raw doublet scores from doublet finder, pegasus, or scrublet, are not included in this release.

--

Each public_id has limited demographic information based on publicly available information in the MMRF CoMMpass study. 

• sex
• Patient sex (not self-identified). Male or Female
• race
• Patient self-identified race
• ethnicity
• Patient self-identified ethnicity
• age
• Patient age at diagnosis. 
• Not reported for patients above 90 at diagnosis
• bmi
• Patient BMI at diagnosis

--

Other

ACCESS TO ADDITIONAL CLINICAL METADATA AND ORIGINAL FASTQ FILES:

All the single-cell raw data, and clinical information used for this analysis, is available at MMRF’s VLAB shared resource (https://mmrfvirtuallab.org) under controlled access to protect patient genetic and clinical information. Access can be requested by filling the VLAB data access request form on the MMRF VLAB website. This form will require basic information such as name, email address, organization, a statement of intended use, and the types of data to be accessed. Senior investigators may apply for access, but must be permanent employees of their institution at a level equivalent to a tenure-track professor. Senior investigators meeting this criteria may nominate members of their immediate team for access. Once a form is submitted, the MMRF data access committee will review the form and respond within one business week from time of submission. If access to the requested data is granted, a temporary link to login and access the dataset will be provided with a one week expiry time under a data transfer agreement that will protect the identities of patients involved in the study. The MMRF VLAB will contain both the processed Seurat object for the original discovery cohort and the combined discovery and validation cohorts, along with the clinical metadata used for samples in the Immune Atlas study. The original unprocessed fastq.gz files for all samples in the Immune Atlas study will also be available.