Oops, you're using an old version of your browser so some of the features on this page may not be displaying properly.

MINIMAL Requirements: Google Chrome 24+Mozilla Firefox 20+Internet Explorer 11Opera 15–18Apple Safari 7SeaMonkey 2.15-2.23

E-Poster Display

896P - Analysis of DLBCL genetic databases using machine learning

Date

17 Sep 2020

Session

E-Poster Display

Topics

Tumour Site

Lymphomas

Presenters

Silvia Sequero Lopez

Citation

Annals of Oncology (2020) 31 (suppl_4): S590-S598. 10.1016/annonc/annonc261

Authors

S. Sequero Lopez1, J. Guarino2, D. Castillo-Barnes3, J.M. Jurado García1, J. Ramirez3, D. Salas-Gonzalez3, J.M. Gorriz3

Author affiliations

  • 1 Medical Oncology Department, Hospital Clinico San Cecilio, 18016 - Granada/ES
  • 2 Department Of Signal Theory, Networking And Communications, University of Granada., 18016 - Granada/ES
  • 3 Department Of Signal Theory, Networking And Communications, University of Granada., 18071 - Granada/ES

Resources

Login to get immediate access to this content.

If you do not have an ESMO account, please create one for free.

Abstract 896P

Background

Diffuse Large B-Cell Lymphoma (DLBCL) is the most frequent type of Non-Hodgkin Lymphoma (NHL). Up to 40% of patients will relapse in the following years with a poor prognosis. Overall Survival (OS) of patients with DLBCL has a wide range depending on the stage and response of the patient. Molecular characteristics could influence the course of DLBCL. Until now, few relevant genetic alterations were known. This is the case of translocations in BCL-2, BCL-6 and MYC whose worse prognosis is a well known fact although they do not change the treatment. The inclusion of modern genetic sequencing techniques have led researchers to find new evidences about DLBCL.

Methods

In this work, two DLBCL genetic datasets available at cbioportal.org have been evaluated to discern between patients with a large OS and subjects with worse prognosis: Duke Cell 2017 dataset which includes 1001 subjects and 150 genetic drivers; and MD Anderson 2013 dataset which contains 23109 mutated genes from 148 participants. To determine if the copy number of these genetic alterations are related to the OS, 4 classification experiments were defined by considering different OS thresholds (12, 24, 36 and 48 months). For example, the first experiment compared subjects with OS < 12 months versus subjects with OS > 12 months.

Results

All the comparisons were performed using a Machine Learning schema composed by: a 10-fold cross-validation loop, a feature selection based on ANOVA (selecting only these values with p-values < 0.01) and a Support-Vector-Machine classifier with Linear Kernel. If classification results were high enough (80% of balanced accuracy is used as reference), then it could be stated that significative genes or their combination are indicative of a better/worse OS. Except by OS threshold of 48 months when using Duke Cell 2017 dataset (54.24%), all the experiments involving MD Anderson 2013 dataset resulted in better classification rates in terms of their balanced accuracy: 59.79% (12 months); 55.75% (24 months); 52.25% (36 months). Nevertheless, all these results were far away of our proposed 80% reference level even despite having p-values < 0.01.

Conclusions

Based on these results, it can be stated that neither the copy number of the genetic alterations nor their combination are related to a greater/lesser OS.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.

This site uses cookies. Some of these cookies are essential, while others help us improve your experience by providing insights into how the site is being used.

For more detailed information on the cookies we use, please check our Privacy Policy.

Customise settings
  • Necessary cookies enable core functionality. The website cannot function properly without these cookies, and you can only disable them by changing your browser preferences.