Cancer Cells Burn Energy on Futile Transcription by Shifting Their Genetic Start Lines

Cancer has long been described as a disease of runaway growth, but a new study suggests it is also a disease of wasteful reading. Researchers at the Indian Institute of Technology Gandhinagar report that tumor cells systematically abandon an energy-conserving design built into the human genome, choosing transcription start sites that force the cell to transcribe long stretches of DNA for no benefit. The work, published in Frontiers in Bioinformatics, reframes a familiar cancer trait — its ravenous appetite for energy — as something written directly into the geometry of where genes begin.

The team focused on a subtlety of gene architecture that is often overlooked. Most human protein-coding genes do not have a single transcription start site, or TSS, but several: the analysis uncovered 57,461 start sites mapped to 16,302 genes, meaning roughly three options per gene, with more than three-quarters of genes carrying at least two. Because these alternative start sites usually sit upstream of the same start codon, switching between them typically leaves the final protein untouched. Yet every nucleotide transcribed between the start site and the codon costs the cell phosphate bonds, so the choice of where RNA polymerase fires is, in energy terms, anything but free.

The researchers call the region between a transcription start site and the start codon a measure of futile transcription distance. When a cell uses a start site very close to the codon, the wasted stretch is short. When it fires from a far-flung upstream site, it transcribes thousands of extra bases — including introns — that code for nothing. Strikingly, the data showed that the human genome has evolved to keep most start sites tightly clustered: about 60 percent of all mapped TSSs lie within one kilobase of their start codon, and successive start sites within a gene showed strong constraints on how much farther apart they could drift. The authors interpret this clustering as the footprint of evolutionary pressure to minimize pointless transcription.

To test whether cancer disrupts this design, the team mined cap analysis of gene expression (CAGE) data from the FANTOM5 project, comparing 476 cancer samples with 1,255 non-cancer samples, and then validated the results using an independent technique called PRO-cap on 110 cancer and 305 non-cancer samples. CAGE and PRO-cap both capture the capped 5′ ends of freshly initiated RNA molecules, providing a direct readout of how often each start site is deployed. By combining these deployment frequencies with the distances to each gene’s start codon, the researchers computed an energy cost coefficient for every gene and, ultimately, for entire transcriptomes.

The result was unambiguous: the aggregated energy cost of the cancer promoterome was roughly 10 to the power of 0.5 — a little over threefold — higher than that of non-cancer samples, an excess confirmed by both technologies after correction for sequencing depth. At the level of individual genes, the aggregated cost coefficient was significantly higher in cancers across 15,861 genes. In other words, cancer cells do not merely transcribe more; they transcribe more wastefully, pushing polymerase to fire from start sites that saddle the cell with longer, more expensive upstream transcripts.

But the deregulation proved far more strategic than simple sloppiness. Using Shannon entropy to quantify how randomly start sites were deployed, the researchers found two contrasting behaviors coexisting in cancer. For most genes, particularly those at distant upstream positions, start site deployment became highly stochastic — scattered, unregulated, and gene-function agnostic, effectively leaking transcription from sites that healthy cells keep silent. Yet for a specific subset of genes, cancers did the opposite: deployment became more deterministic, tightly concentrated on the most proximal, energy-cheap start sites. That subset was overwhelmingly enriched for genes driving biosynthesis, nucleotide metabolism, cell proliferation, and the cell cycle — the very engines of the cancer phenotype.

This duality, the authors argue, reflects selection acting on energy economics. A gene that must be expressed at high volume in a rapidly dividing cell profits enormously from shaving thousands of wasted bases off every transcript. The genome apparently anticipated this: biosynthetic and proliferation genes already carry start sites that sit unusually close to their start codons, a layout the researchers suggest may have first evolved to support the rapid cell divisions of early embryogenesis under metabolic constraint. Incipient cancer cells appear to co-opt this pre-existing economy, and even recruit chromatin-regulatory genes — whose deregulated deployment could rewire the epigenome — to lock in the cost-saving configuration.

The study also proposes an evolutionary interpretation of transcriptional noise itself. If start sites are clustered proximally, then purely random deployment of them actually saves energy, because the dice fall most often on the short transcripts. The authors suggest the human transcription start site landscape has been shaped to make stochasticity cheap, and that healthy cells exploit this, maintaining higher gene-space entropy with lower overall cost. Cancer breaches that threshold in two directions at once: embracing expensive randomness at distant start sites while carving out tightly regulated, low-cost proximal deployment for hallmark genes. The combined parameter the team built from effective futile transcription distance and compound entropy turned out to be the single strongest classifier of cancer from non-cancer samples in their functional analyses.

The findings carry caveats the authors themselves acknowledge. The observed patterns could partly be downstream consequences of MYC-driven transcriptional amplification, elevated RNA polymerase II activity, or the metabolic reprogramming of specific tumor types, rather than a truly universal hallmark — and the magnitude of the cost excess varied across tissue types, with brain cancers a notable exception in the CAGE data. Even so, the near-universal direction of the effect across heterogeneous sample pools, replicated across two independent technologies, suggests that transcriptional energy waste deserves a place alongside the Warburg effect and glutamine addiction as a metabolic signature of malignancy.

Beyond cancer biology, the work hints at a general principle: where a gene begins is not an accident of genome annotation but a calibrated trade-off between regulatory flexibility and metabolic cost. If future experiments confirm the causal chain — that relaxing start site selection precedes and enables malignant transformation — then the transcriptional start site landscape could become a diagnostic window onto a tumor’s energetic state, and perhaps even a therapeutic target. For now, the message is vivid enough: in cancer, even the silent stretches of RNA carry a bill, and the cell has learned to make someone else pay it.

Subject of Research: Deregulated transcription start site deployment and futile transcription energetics in cancer

Article Title: Gene functions determine stochastic or adaptive futile transcription in cancers through deregulated start site deployments

Article References: Satish, A. L., Kumar, P., Morbia, I., & Singh, U. (2026). Gene functions determine stochastic or adaptive futile transcription in cancers through deregulated start site deployments. Frontiers in Bioinformatics, 6, Article 1883461. https://doi.org/10.3389/fbinf.2026.1883461

Image Credits: AI Generated

DOI: 10.3389/fbinf.2026.1883461

Keywords: transcription start sites, futile transcription, cancer metabolism, CAGE, PRO-cap, gene expression, Shannon entropy, promoterome, Warburg effect, energy cost, biosynthesis, cell proliferation