It has been observed that both cancer tissue cells and normal proliferating cells (NPCs) have the Warburg effect. Our goal here is to demonstrate that they do this for different reasons. To accomplish this, we have analyzed the transcriptomic data of over 7000 cancer and control tissues of 14 cancer types in TCGA and data of five NPC types in GEO. Our analyses reveal that NPCs accumulate large quantities of ATPs produced by the respiration process before starting the Warburg effect, to raise the intracellular pH from ∼6.8 to ∼7.2 and to prepare for cell division energetically. Once cell cycle starts, the cells start to rely on glycolysis for ATP generation followed by ATP hydrolysis and lactic acid release, to maintain the elevated intracellular pH as needed by cell division since together the three processes are pH neutral. The cells go back to the normal respiration-based ATP production once the cell division phase ends. In comparison, cancer cells have reached their intracellular pH at ∼7.4 from top down as multiple acid-loading transporters are up-regulated and most acid-extruding ones except for lactic acid exporters are repressed. Cancer cells use continuous glycolysis for ATP production as way to acidify the intracellular space since the lactic acid secretion is decoupled from glycolysis-based ATP generation and is pH balanced by increased expressions of acid-loading transporters. Co-expression analyses suggest that lactic acid secretion is regulated by external, non-pH related signals. Overall, our data strongly suggest that the two cell types have the Warburg effect for very different reasons.

Considerable metabolic reprogramming has been observed in a conserved manner across multiple cancer types, but their true causes remain elusive. We present an analysis of around 50 such reprogrammed metabolisms (RM) including the Warburg effect, nucleotide de novo synthesis, and sialic acid biosynthesis in cancer. Analyses of the biochemical reactions conducted by these RMs, coupled with gene expression data of their catalyzing enzymes, in 7,011 tissues of 14 cancer types, revealed that all RMs produce more H+ than their original metabolisms. These data strongly support a model that these RMs are induced or selected to neutralize a persistent intracellular alkaline stress due to chronic inflammation and local iron overload. To sustain these RMs for survival, cells must find metabolic exits for the nonproton products of these RMs in a continuous manner, some of which pose major challenges, such as nucleotides and sialic acids, because they are electrically charged. This analysis strongly suggests that continuous cell division and other cancerous behaviors are ways for the affected cells to remove such products in a timely and sustained manner. As supporting evidence, this model can offer simple and natural explanations to a range of long-standing open questions in cancer research including the cause of the Warburg effect. Significance: Inhibiting acidifying metabolic reprogramming could be a novel strategy for treating cancer.

We present a computational study of tissue transcriptomic data of 14 cancer types to address: what may drive cancer cell division? Our analyses point to that persistent disruption of the intracellular pH by Fenton reactions may be at the root of cancer development. Specifically, we have statistically demonstrated that Fenton reactions take place in cancer cytosol and mitochondria across all the 14 cancer types, based on cancer tissue gene-expression data integrated via the Michaelis–Menten equation. In addition, we have shown that (i) Fenton reactions in cytosol of the disease cells will continuously increase their pH, to which the cells respond by generating net protons to keep the pH stable through a combination of synthesizing glycolytic ATPs and consuming them by nucleotide syntheses, which may drive cell division to rid of the continuously synthesized nucleotides; and (ii) Fenton reactions in mitochondria give rise to novel ways for ATP synthesis with electrons ultimately coming from H₂O₂, largely originated from immune cells. A model is developed to link these to cancer development, where some mutations may be selected to facilitate cell division at rates dictated by Fenton reactions.

Accurate molecular subtypes prediction of cancer patients is significant for personalized cancer diagnosis and treatments. Large amount of multi-omics data and the advancement of data-driven methods are expected to facilitate molecular subtyping of cancer. Most existing machine learning–based methods usually classify samples according to single omics data, fail to integrate multi-omics data to learn comprehensive representations of the samples, and ignore that information transfer and aggregation among samples can better represent them and ultimately help in classification. We propose a novel framework named multi-omics graph convolutional network (M-GCN) for molecular subtyping based on robust graph convolutional networks integrating multi-omics data. We first apply the Hilbert–Schmidt independence criterion least absolute shrinkage and selection operator (HSIC Lasso) to select the molecular subtype-related transcriptomic features and then construct a sample–sample similarity graph with low noise by using these features. Next, we take the selected gene expression, single nucleotide variants (SNV), and copy number variation (CNV) data as input and learn the multi-view representations of samples. On this basis, a robust variant of graph convolutional network (GCN) model is finally developed to obtain samples’ new representations by aggregating their subgraphs. Experimental results of breast and stomach cancer demonstrate that the classification performance of M-GCN is superior to other existing methods. Moreover, the identified subtype-specific biomarkers are highly consistent with current clinical understanding and promising to assist accurate diagnosis and targeted drug development.

Gene regulatory network (GRN) inference from gene expression data is a significant approach to understanding aspects of the biological system. Compared with generalized correlation‐based methods, causality‐inspired ones seem more rational to infer regulatory relationships. We propose GRINCD, a novel GRN inference framework empowered by graph representation learning and causal asymmetric learning, considering both linear and non‐linear regulatory relationships. First, high‐quality representation of each gene is generated using graph neural network. Then, we apply the additive noise model to predict the causal regulation of each regulator‐target pair. Additionally, we design two channels and finally assemble them for robust prediction. Through comprehensive comparisons of our framework with state‐of‐the‐art methods based on different principles on numerous datasets of diverse types and scales, the experimental results show that our framework achieves superior or comparable performance under various evaluation metrics. Our work provides a new clue for constructing GRNs, and our proposed framework GRINCD also shows potential in identifying key factors affecting cancer development.

Cancer is a heterogeneous disease that is driven by the accumulation of both genetic and nongenetic alterations, so integrating multiomics data and extracting effective information from them is expected to be an effective way to predict cancer driver genes. In this paper, we first generate comprehensive instructive features for each gene from genomic, epigenomic, transcriptomic levels together with protein–protein interaction (PPI)-networks-derived attributes and then propose a novel semisupervised deep graph learning framework GGraphSAGE to predict cancer driver genes according to the impact of the alterations on a biological system. When applied to eight tumor types, experimental results suggest that GGraphSAGE outperforms several state-of-the-art computational methods for driver genes identification. Moreover, it broadens our current understanding of cancer driver genes from multiomics level and identifies driver genes specific to the tumor type rather than pan-cancer. We expect GGraphSAGE to open new avenues in precision medicine and even further predict drivers for other complex diseases.