The Paper That Will Never Be Published
In the top drawer of my wooden desk sits a manila folder holding thirty-two pages of printed figures, statistical tables, and half-written paragraphs. The title at the top of page one is ambitious, elegant, and completely wrong.
For eight months in 2022, I was convinced that a specific long non-coding RNA acted as a master competitive endogenous RNA (ceRNA) “sponge” for miR-21 in colorectal adenocarcinoma. The initial correlation in public TCGA data had an impressive Pearson coefficient (). The preliminary target predictions in TargetScan were flawless. The story was so compelling that the introduction wrote itself in a single weekend.
THE CE-RNA HYPOTHESIS THAT FAILED
[lncRNA sponge] ──────(Sequesters miR-21)──────> [PTEN preserved]
│ │
(Downregulated) (Downregulated)
│ │
▼ ▼
Free miR-21 binds PTEN 3'UTR ──────────────> Aggressive Tumor Phenotype
Then came the laboratory knockdowns and synthetic reporter assays.
No matter how high we overexpressed the lncRNA, miR-21 activity barely budged. When we knocked it down with antisense oligonucleotides, PTEN mRNA expression remained identical to the scramble control. The stoichiometry simply did not work: there were roughly 15,000 copies of miR-21 per cell, but only 40 copies of the putative lncRNA sponge. You cannot mop up a flooded cellar with a dry postage stamp.
The hypothesis was dead.
The File-Drawer Problem and Scientific Bias
In academic publishing, there is immense pressure to convert every endeavor into a clean, triumphant linear narrative: question asked, clever method applied, breathtaking discovery confirmed.
This selective reporting creates the well-documented file-drawer effect: Thousands of labs worldwide repeatedly test the same plausible yet incorrect biological ideas, spend millions of grant dollars, hit the exact same stoichiometric impossibility, and quietly bury the result in a folder just like mine. Because negative results are difficult to publish in high-impact venues, the wider scientific community never learns that the path leads to a cliff.
What Actually Happened What Journals Prefer
┌─────────────────────────────┐ ┌─────────────────────────────┐
│ 1. Hypothesis A (Failed) │ │ 1. Polished Retrospective │
│ 2. Stoichiometric check (No)│ ─────────> │ Hypothesis │
│ 3. 6 Months of negative data│ │ 2. Clear Linear Success │
│ 4. Hypothesis revised │ │ 3. P < 0.01 Confirmed │
│ 5. Folder filed away │ │ │
└─────────────────────────────┘ └─────────────────────────────┘
What the Folder Taught Me
Setting aside that manuscript was painful. It meant admitting that weeks of computational pipelines and wet-lab hours had produced no claim worthy of a press release.
Yet looking back, that unfinished project was among the most formative experiences of my career. It taught me:
- Check the numbers before falling in love with a mechanism: Biological plausibility is worthless if molecule counts differ by three orders of magnitude.
- Correlation is not even causation’s cousin: High correlation across 500 patient biopsies frequently reflects passenger effects or common cell-type proportion shifts, not direct molecular interaction.
- Grace in letting go: Tenacity in science is commendable; stubborn refusal to accept negative evidence is pathology.
The manuscript remains in the drawer. Occasionally, when a new algorithm or exciting correlation seems too neat to be true, I pull it out and reread the opening paragraph. It serves as an indispensable reminder that science moves forward not only by proving what is, but by humbly documenting what is not.
Further Reading
- Matosin, N. et al. Negativity towards negative results: A perspective. Dis. Model. Mech. 7, 171–173 (2014).
- Knight, J. Negative results: Null and alternative hypotheses. Nature 422, 545 (2003).
