Developing Robust Predictive Models for Head and Neck Cancer across Microarray and RNA-seq Data

Deep learning based feature-level integration of multi-omics data for breast cancer patients survival analysis

BACKGROUND

Breast cancer is the most prevalent and among the most deadly cancers in females. Patients with breast cancer have highly variable survival lengths, indicating a need to identify prognostic biomarkers for personalized diagnosis and treatment. With the development of new technologies such as next-generation sequencing, multi-omics information are becoming available for a more thorough evaluation of a patient's condition. In this study, we aim to improve breast cancer overall survival prediction by integrating multi-omics data (e.g., gene expression, DNA methylation, miRNA expression, and copy number variations (CNVs)).

METHODS

Motivated by multi-view learning, we propose a novel strategy to integrate multi-omics data for breast cancer survival prediction by applying complementary and consensus principles. The complementary principle assumes each -omics data contains modality-unique information. To preserve such information, we develop a concatenation autoencoder (ConcatAE) that concatenates the hidden features learned from each modality for integration. The consensus principle assumes that the disagreements among modalities upper bound the model errors. To get rid of the noises or discrepancies among modalities, we develop a cross-modality autoencoder (CrossAE) to maximize the agreement among modalities to achieve a modality-invariant representation. We first validate the effectiveness of our proposed models on the MNIST simulated data. We then apply these models to the TCCA breast cancer multi-omics data for overall survival prediction.

RESULTS

For breast cancer overall survival prediction, the integration of DNA methylation and miRNA expression achieves the best overall performance of 0.641 ± 0.031 with ConcatAE, and 0.63 ± 0.081 with CrossAE. Both strategies outperform baseline single-modality models using only DNA methylation (0.583 ± 0.058) or miRNA expression (0.616 ± 0.057).

CONCLUSIONS

In conclusion, we achieve improved overall survival prediction performance by utilizing either the complementary or consensus information among multi-omics data. The proposed ConcatAE and CrossAE models can inspire future deep representation-based multi-omics integration techniques. We believe these novel multi-omics integration models can benefit the personalized diagnosis and treatment of breast cancer patients.

Integrating multi-omics data by learning modality invariant representations for improved prediction of overall survival of cancer

Breast and ovarian cancers are the second and the fifth leading causes of cancer death among women. Predicting the overall survival of breast and ovarian cancer patients can facilitate the therapeutics evaluation and treatment decision making. Multi-scale multi-omics data such as gene expression, DNA methylation, miRNA expression, and copy number variations can provide insights on personalized survival. However, how to effectively integrate multi-omics data remains a challenging task. In this paper, we develop multi-omics integration methods to improve the prediction of overall survival for breast cancer and ovarian cancer patients. Because multi-omics data for the same patient jointly impact the survival of cancer patients, features from different -omics modality are related and can be modeled by either association or causal relationship (e.g., pathways). By extracting these relationships among modalities, we can get rid of the irrelevant information from high-throughput multi-omics data. However, it is infeasible to use the Brute Force method to capture all possible multi-omics interactions. Thus, we use deep neural networks with novel divergence-based consensus regularization to capture multi-omics interactions implicitly by extracting modality-invariant representations. In comparing the concatenation-based integration networks with our new divergence-based consensus networks, the breast cancer overall survival C-index is improved from 0.655±0.062 to 0.671±0.046 when combing DNA methylation and miRNA expression, and from 0.627±0.062 to 0.667±0.073 when combing miRNA expression and copy number variations. In summary, our novel deep consensus neural network has successfully improved the prediction of overall survival for breast cancer and ovarian cancer patients by implicitly learning the multi-omics interactions.

QuantSeq. 3' Sequencing combined with Salmon provides a fast, reliable approach for high throughput RNA expression analysis

RNA-Seq is increasingly used for the diagnosis of patients, targeting of therapies and for single cell transcriptomics. These applications require cost effective, fast and reliable ways of capturing and analyzing gene expression data. Here we compared Lexogen's QuantSeq which captures only the 3' end of RNA transcripts and Illumina's TruSeq, using both Tophat2 and Salmon for gene quantification. We also compared these results to microarray. This analysis was performed on peripheral blood mononuclear cells stimulated with Poly (I:C), a viral mimic that induces innate antiviral responses. This provides a well-established model to determine if RNA-Seq and QuantSeq identify the same biological signatures. Gene expression levels in QuantSeq and RNA-Seq were strongly correlated (Spearman's rho ~0.8), Salmon and Tophat2 (Spearman's rho > 0.9). There was high consistency in protein coding genes, non-concordant genes had a high proportion of shorter, non-coding features. RNA-Seq identified more differentially expressed genes than QuantSeq, both methods outperformed microarray. The same key biological signals emerged in each of these approaches. We conclude that QuantSeq, coupled with a fast quantification method such as Salmon, should provide a viable alternative to traditional RNA-Seq in many applications and may be of particular value in the study of the 3'UTR region of mRNA.

Tumor cell sensitivity to vemurafenib can be predicted from protein expression in a BRAF-V600E basket trial setting

BACKGROUND

Genetics-based basket trials have emerged to test targeted therapeutics across multiple cancer types. However, while vemurafenib is FDA-approved for BRAF-V600E melanomas, the non-melanoma basket trial was unsuccessful, suggesting mutation status is insufficient to predict response. We hypothesized that proteomic data would complement mutation status to identify vemurafenib-sensitive tumors and effective co-treatments for BRAF-V600E tumors with inherent resistance.

METHODS

Reverse Phase Proteomic Array (RPPA, MD Anderson Cell Lines Project), RNAseq (Cancer Cell Line Encyclopedia) and vemurafenib sensitivity (Cancer Therapeutic Response Portal) data for BRAF-V600E cancer cell lines were curated. Linear and nonlinear regression models using RPPA protein or RNAseq were evaluated and compared based on their ability to predict BRAF-V600E cell line sensitivity (area under the dose response curve). Accuracies of all models were evaluated using hold-out testing. CausalPath software was used to identify protein-protein interaction networks that could explain differential protein expression in resistant cells. Human examination of features employed by the model, the identified protein interaction networks, and model simulation suggested anti-ErbB co-therapy would counter intrinsic resistance to vemurafenib. To validate this potential co-therapy, cell lines were treated with vemurafenib and dacomitinib (a pan-ErbB inhibitor) and the number of viable cells was measured.

RESULTS

Orthogonal partial least squares (O-PLS) predicted vemurafenib sensitivity with greater accuracy in both melanoma and non-melanoma BRAF-V600E cell lines than other leading machine learning methods, specifically Random Forests, Support Vector Regression (linear and quadratic kernels) and LASSO-penalized regression. Additionally, use of transcriptomic in place of proteomic data weakened model performance. Model analysis revealed that resistant lines had elevated expression and activation of ErbB receptors, suggesting ErbB inhibition could improve vemurafenib response. As predicted, experimental evaluation of vemurafenib plus dacomitinb demonstrated improved efficacy relative to monotherapies.

CONCLUSIONS

Combined, our results support that inclusion of proteomics can predict drug response and identify co-therapies in a basket setting.