Dataset Information

CTULIP: application of a human-based RNA-seq primary tumor classification tool for cross-species primary tumor classification in canine.

ABSTRACT:

Introduction

The domestic dog, Canis familiaris, is quickly gaining traction as an advantageous model for use in the study of cancer, one of the leading causes of death worldwide. Naturally occurring canine cancers share clinical, histological, and molecular characteristics with the corresponding human diseases.

Methods

In this study, we take a deep-learning approach to test how similar the gene expression profile of canine glioma and bladder cancer (BLCA) tumors are to the corresponding human tumors. We likewise develop a tool for identifying misclassified or outlier samples in large canine oncological datasets, analogous to that which was developed for human datasets.

Results

We test a number of machine learning algorithms and found that a convolutional neural network outperformed logistic regression and random forest approaches. We use a recently developed RNA-seq-based convolutional neural network, TULIP, to test the robustness of a human-data-trained primary tumor classification tool on cross-species primary tumor prediction. Our study ultimately highlights the molecular similarities between canine and human BLCA and glioma tumors, showing that protein-coding one-to-one homologs shared between humans and canines, are sufficient to distinguish between BLCA and gliomas.

Discussion

The results of this study indicate that using protein-coding one-to-one homologs as the features in the input layer of TULIP performs good primary tumor prediction in both humans and canines. Furthermore, our analysis shows that our selected features also contain the majority of features with known clinical relevance in BLCA and gliomas. Our success in using a human-data-trained model for cross-species primary tumor prediction also sheds light on the conservation of oncological pathways in humans and canines, further underscoring the importance of the canine model system in the study of human disease.

SUBMITTER: Long J

PROVIDER: S-EPMC10397722 | biostudies-literature | 2023

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

cTULIP: application of a human-based RNA-seq primary tumor classification tool for cross-species primary tumor classification in canine.

Long Jiaxin J Ganakammal Satishkumar Ranganathan SR Jones Sara E SE Kothandaraman Harish H Dhawan Deepika D Ogas Joe J Knapp Deborah W DW Beyers Matthew M Lanman Nadia A NA

Frontiers in oncology 20230720

<h4>Introduction</h4>The domestic dog, <i>Canis familiaris</i>, is quickly gaining traction as an advantageous model for use in the study of cancer, one of the leading causes of death worldwide. Naturally occurring canine cancers share clinical, histological, and molecular characteristics with the corresponding human diseases.<h4>Methods</h4>In this study, we take a deep-learning approach to test how similar the gene expression profile of canine glioma and bladder cancer (BLCA) tumors are to the ...[more]

PMID: 37546395

Similar Datasets

Project description:We report transcriptomes from 430 single glioblastoma cells isolated from 5 individual tumors and 102 single cells from gliomasphere cells lines generated using SMART-seq. In addition, we report population RNA-seq from the five tumors as well as RNA-seq from cell lines derived from 3 tumors (MGH26, MGH28, MGH31) cultured under serum free (CSC) and differentiated (FCS) conditions. This dataset highlights intratumoral heterogeneity with regards to the expression of de novo derived transcriptional modules and established subtype classifiers. Operative specimens from five glioblastoma patients (MGH26, MGH28, MGH29, MGH30, MGH31) were acutely dissociated, depleted for CD45+ inflammatory cells and then sorted as single cells (576 samples). Population controls for each tumor were isolated by sorting 2000-10000 cells and processed in parallel (5 population control samples). Single cells from two established cell lines, GBM6 and GBM8, were also sorted as single cells (192 samples). SMART-seq protocol was implemented to generate single cell full length transcriptomes (modified from Shalek, et al Nature 2013) and sequenced using 25 bp paired end reads. Single cell cDNA libraries for MGH30 were resequenced using 100 bp paired end reads to allow for isoform and splice junction reconstruction (96 samples, annotated MGH30L). Cells were also cultured in serum free conditions to generate gliomasphere cell lines for MGH26, MGH28, and MGH31 (CSC) which were then differentiated using 10% serum (FCS). Population RNA-seq was performed on these samples (3 CSC, 3 FCS, 6 total). The initial dataset included 875 RNA-seq libraries (576 single glioblastoma cells, 96 resequenced MGH30L, 192 single gliomasphere cells, 5 tumor population controls, 6 population libraries from CSC and FCS samples). Data was processed as described below using RSEM for quantification of gene expression. 5,948 genes with the highest composite expression either across all single cells combined (average log2(TPM)>4.5) or within a single tumor (average log2(TPM)>6 in at least one tumor) were included. Cells expressing less than 2,000 of these 5,948 genes were excluded. The final processed dataset then included 430 primary single cell glioblastoma transcriptomes, 102 single cell transcriptomes from cell lines(GBM6,GBM8), 5 population controls (1 for each tumor), and 6 population libraries from cell lines derived from the tumors (CSC and FCS for MGH26, MGH28 and MGH31). The final matrix (GBM_data_matrix.txt) therefore contains 5948 rows (genes) quantified in 543 samples (columns). Please note that the samples which are not included in the data processing are indicated in the sample description field.