Identifying the oncogenic potential of gene fusions exploiting miRNAs

Pathogenicity Prediction of Gene Fusion in Structural Variations: A Knowledge Graph-Infused Explainable Artificial Intelligence (XAI) Framework

When analyzing cancer sample genomes in clinical practice, many structural variants (SVs), other than single nucleotide variants (SNVs), have been identified. To identify driver variants, the leading candidates must be narrowed down. When fusion genes are involved, selection is particularly difficult, and highly accurate predictions from AI is important. Furthermore, we also wanted to determine how the prediction can make more reliable diagnoses. Here, we developed an explainable AI (XAI) suitable for SVs with gene fusions, based on the XAI technology we previously developed for the prediction of SNV pathogenicity. To cope with gene fusion variants, we added new data to the previous knowledge graph for SVs and we improved the algorithm. Its prediction accuracy was as high as that of existing tools. Moreover, our XAI could explain the reasons for these predictions. We used some variant examples to demonstrate that the reasons are plausible in terms of pathogenic basic mechanisms. These results can be seen as a hopeful step toward the future of genomic medicine, where efficient and correct decisions can be made with the support of AI.

Integrated microRNA and proteome analysis of cancer datasets with MoPC

MicroRNAs (miRNAs) are small molecules that play an essential role in regulating gene expression by post-transcriptional gene silencing. Their study is crucial in revealing the fundamental processes underlying pathologies and, in particular, cancer. To date, most studies on miRNA regulation consider the effect of specific miRNAs on specific target mRNAs, providing wet-lab validation. However, few tools have been developed to explain the miRNA-mediated regulation at the protein level. In this paper, the MoPC computational tool is presented, that relies on the partial correlation between mRNAs and proteins conditioned on the miRNA expression to predict miRNA-target interactions in multi-omic datasets. MoPC returns the list of significant miRNA-target interactions and plot the significant correlations on the heatmap in which the miRNAs and targets are ordered by the chromosomal location. The software was applied on three TCGA/CPTAC datasets (breast, glioblastoma, and lung cancer), returning enriched results in three independent targets databases.

Exploring the Relationship between Fusion Genes and MicroRNAs in Cancer

Fusion genes are key cancer driver genes that can be used as potential drug targets in precision therapies, and they can also serve as accurate diagnostic and prognostic biomarkers. The fusion genes can cause microRNA (miRNA/miR) aberrations in many types of cancer. Nevertheless, whether fusion genes incite miRNA aberrations as one of their many critical oncogenic functionalities for driving carcinogenesis needs further investigation. Recent discoveries of miRNA genes that are present within the regions of genomic rearrangements that initiate fusion gene-based intronic miRNA dysregulation have brought the fusion genes into the limelight and revealed their unexplored potential in the field of cancer biology. Fusion gene-based 'promoter-switch' event aberrantly activate the miRNA-related upstream regulatory signals, while fusion-based coding region alterations disrupt the original miRNA coding loci. Fusion genes can potentially regulate the miRNA aberrations regardless of the protein-coding capability of the resultant fusion transcript. Studies on out-of-frame fusion and nonrecurrent fusion genes that cause miRNA dysregulation have attracted the attention of researchers on fusion genes from an oncological perspective and therefore could have potential implications in cancer therapies. This review will provide insights into the role of fusion genes and miRNAs, and their possible interrelationships in cancer.

Predicting gene and protein expression levels from DNA and protein sequences with Perceiver

BACKGROUND AND OBJECTIVE

The functions of an organism and its biological processes result from the expression of genes and proteins. Therefore quantifying and predicting mRNA and protein levels is a crucial aspect of scientific research. Concerning the prediction of mRNA levels, the available approaches use the sequence upstream and downstream of the Transcription Start Site (TSS) as input to neural networks. The State-of-the-art models (e.g., Xpresso and Basenjii) predict mRNA levels exploiting Convolutional (CNN) or Long Short Term Memory (LSTM) Networks. However, CNN prediction depends on convolutional kernel size, and LSTM suffers from capturing long-range dependencies in the sequence. Concerning the prediction of protein levels, as far as we know, there is no model for predicting protein levels by exploiting the gene or protein sequences.

METHODS

Here, we exploit a new model type (called Perceiver) for mRNA and protein level prediction, exploiting a Transformer-based architecture with an attention module to attend to long-range interactions in the sequences. In addition, the Perceiver model overcomes the quadratic complexity of the standard Transformer architectures. This work's contributions are 1. DNAPerceiver model to predict mRNA levels from the sequence upstream and downstream of the TSS; 2. ProteinPerceiver model to predict protein levels from the protein sequence; 3. Protein&DNAPerceiver model to predict protein levels from TSS and protein sequences.

RESULTS

The models are evaluated on cell lines, mice, glioblastoma, and lung cancer tissues. The results show the effectiveness of the Perceiver-type models in predicting mRNA and protein levels.

CONCLUSIONS

This paper presents a Perceiver architecture for mRNA and protein level prediction. In the future, inserting regulatory and epigenetic information into the model could improve mRNA and protein level predictions. The source code is freely available at https://github.com/MatteoStefanini/DNAPerceiver.

In silico validation of RNA-Seq results can identify gene fusions with oncogenic potential in glioblastoma

RNA-Sequencing (RNA-Seq) can identify gene fusions in tumors, but not all these fusions have functional consequences. Using multiple data bases, we have performed an in silico analysis of fusions detected by RNA-Seq in tumor samples from 139 newly diagnosed glioblastoma patients to identify in-frame fusions with predictable oncogenic potential. Among 61 samples with fusions, there were 103 different fusions, involving 167 different genes, including 20 known oncogenes or tumor suppressor genes (TSGs), 16 associated with cancer but not oncogenes or TSGs, and 32 not associated with cancer but previously shown to be involved in fusions in gliomas. After selecting in-frame fusions able to produce a protein product and running Oncofuse, we identified 30 fusions with predictable oncogenic potential and classified them into four non-overlapping categories: six previously described in cancer; six involving an oncogene or TSG; four predicted by Oncofuse to have oncogenic potential; and 14 other in-frame fusions. Only 24 patients harbored one or more of these 30 fusions, and only two fusions were present in more than one patient: FGFR3::TACC3 and EGFR::SEPTIN14. This in silico study provides a good starting point for the identification of gene fusions with functional consequences in the pathogenesis or treatment of glioblastoma.