Description
Contemporary analyses focused on a limited number of clinical and molecular features have been unable to accurately predict clinical outcomes in pancreatic ductal adenocarcinoma (PDAC). Here we describe a novel, conceptual approach and use it to analyze clinical, computational pathology, and molecular (DNA, RNA, protein, and lipid) analyte data from 74 patients with resectable PDAC. Multiple, independent, machine learning models were developed and tested on curated single and multi-omic feature/analyte panels to determine their ability to predict clinical outcomes in patients. The multi-omic models predicted recurrence with an accuracy and positive predictive value (PPV) of 0.90, 0.91, and survival of 0.85, 0.87, respectively, outperforming every single-omic model. In predicting survival, we defined a parsimonious model with only 589 multi-omic analytes that had an accuracy and PPV of 0.85. Our approach enables discovery of parsimonious biomarker panels with similar predictive performance to that of larger and resource consuming panels and thereby has a significant potential to democratize precision cancer medicine worldwide.
[doi:10.25345/C5GM8202W]
[dataset license: CC0 1.0 Universal (CC0 1.0)]
Keywords: multiomics ; lipidomics
Contact
Principal Investigators:
(in alphabetical order)
|
Jennifer Van Eyk, Cedars Sinai Medical Center, United States
|
Submitting User: |
NivedaS5
|
Number of Files: |
|
Total Size: |
|
Spectra: |
|
Subscribers: |
|
|
| Owner |
Conditions:
|
|
|
Biological Replicates:
|
|
|
Technical Replicates:
|
|
|
|
Identification Results |
Proteins (Human, Remapped):
|
|
|
Proteins (Reported):
|
|
|
Peptides:
|
|
|
Variant Peptides:
|
|
|
PSMs:
|
|
|
|
Differential Proteins:
|
|
|
Quantified Proteins:
|
|
|
|
GNPS content goes here (MSV000093118 [task=6ae04d67611f4635ad04d9b4b03b71ec])
Click here to queue conversion of this dataset's submitted spectrum files
to open formats (e.g. mzML). This process may take some time.
When complete, the converted files will be available in the "ccms_peak"
subdirectory of the dataset's FTP space (accessible via the "FTP Download"
link to the right).
Number of distinct conditions across all analyses (original submission and reanalyses)
associated with this dataset.
Distinct condition labels are counted across all files submitted in the "Metadata" category
having a "Condition" column in this dataset.
"N/A" means no results of this type were submitted.
Number of distinct biological replicates across all analyses (original submission and reanalyses)
associated with this dataset.
Distinct replicate labels are counted across all files submitted in the "Metadata" category
having a "BioReplicate" or "Replicate" column in this dataset.
"N/A" means no results of this type were submitted.
Number of distinct technical replicates across all analyses (original submission and reanalyses)
associated with this dataset.
The technical replicate count is defined as the maximum number of times any one distinct
combination of condition and biological replicate was analyzed across all files submitted in the
"Metadata" category. In the case of fractionated experiments, only the first fraction is
considered.
"N/A" means no results of this type were submitted.
Originally identified proteins that were automatically
remapped by MassIVE to proteins in the
SwissProt
human reference database.
"N/A" means no results of this type were submitted.
Number of distinct protein accessions reported across all analyses (original submission and
reanalyses) associated with this dataset.
"N/A" means no results of this type were submitted.
Number of distinct unmodified peptide sequences reported across all analyses (original
submission and reanalyses) associated with this dataset.
"N/A" means no results of this type were submitted.
Number of distinct peptide sequences (including modified variants or peptidoforms) reported
across all analyses (original submission and reanalyses) associated with this dataset.
"N/A" means no results of this type were submitted.
Total number of peptide-spectrum matches (i.e. spectrum identifications) reported across all
analyses (original submission and reanalyses) associated with this dataset.
"N/A" means no results of this type were submitted.
Number of distinct proteins quantified across all analyses (original submission and reanalyses)
associated with this dataset.
Distinct protein accessions are counted across all files submitted in the "Statistical Analysis
of Quantified Analytes" category having a "Protein" column in this dataset.
"N/A" means no results of this type were submitted.
Number of distinct proteins found to be differentially abundant in at least one comparison
across all analyses (original submission and reanalyses) associated with this dataset.
A protein is differentially abundant if its change in abundance across conditions is found
to be statistically significant with an adjusted p-value <= 0.05 and lists no issues associated
with statistical tests for differential abundance.
Distinct protein accessions are counted across all files submitted in the "Statistical Analysis
of Quantified Analytes" category having a "Protein" column in this dataset.
"N/A" means no results of this type were submitted.
This dataset may not contain all raw spectra data as originally deposited in PRIDE.
It has been imported to MassIVE for reanalysis purposes, so its spectra data here may
consist solely of processed peak lists suitable for reanalysis with most software.