Field: Biochemistry, Genetics and Molecular Biology › Molecular BiologyType: Couldn't replicate a published resultArea: stem cell biologyTarget: STAP cells (stimulus-triggered acquisition of pluripotency)Stage: replication
Hypothesis
Cells prepared by the STAP method would be pluripotent and contribute to chimeric embryos, as Obokata et al. (2014) reported.
What failed
Under supervision by RIKEN, the original author prepared STAP cells that were injected into 1,051 embryos. Of the 591 recovered, none showed a significant contribution from the injected cells.
Broken assumption
A simple stress such as acid can turn ordinary body cells into pluripotent stem cells.
Why
The abstract does not give a cause. The key evidence for pluripotency could not be reproduced, even by the original researcher.
Field: Biochemistry, Genetics and Molecular Biology › Molecular BiologyType: Couldn't replicate a published resultArea: neuronal ion channelsTarget: Cav1.2 L-type calcium channel α1 subunitStage: preclinical
Hypothesis
The main pore-forming part of the Cav1.2 calcium channel is cut in the middle, between its second and third domains, as Michailidis et al. (2014) reported.
What failed
Systematic immunoblots of brain tissue and cells with six region-specific antibodies found a full-length form (about 250 kDa) and a shorter form missing about 350 residues at the C-terminal end. Cutting in the middle of the channel was at most minimal in brain tissue.
Broken assumption
The Cav1.2 channel protein in neurons is commonly cut into two halves.
Why
The shorter form comes from losing the end of the protein, not from a cut in the middle.
Field: Medicine › NeurologyType: No effect foundArea: parkinson's diseaseTarget: LRRK2 (wild-type and G2019S)Stage: preclinical
Hypothesis
High levels of LRRK2, especially the G2019S mutant that causes inherited Parkinson's disease, would cause or worsen alpha-synuclein disease in mouse brain neurons.
What failed
High LRRK2 left alpha-synuclein and Tau levels unchanged and did not worsen alpha-synuclein disease in mice that also made high alpha-synuclein. In some lines, high LRRK2 improved motor skills.
Broken assumption
Raising LRRK2 levels in neurons is enough to drive or worsen alpha-synuclein disease.
Why
The abstract does not give a cause. Many neurons tolerate high LRRK2 well.
Field: Biochemistry, Genetics and Molecular Biology › Cancer ResearchType: Couldn't replicate a published resultArea: nutritionTarget: Dietary milk microRNAs in human bloodStage: replication
Hypothesis
Drinking cow's milk would raise blood levels of miR-29b and miR-200c in adults, as a recent report found, possibly because milk microRNAs pass into the blood.
What failed
Re-testing samples from the original study, 223 small RNAs showed no significant change after drinking milk, confirmed by qPCR. Sequencing found no miR-29b or miR-200c reads and no uniquely bovine microRNAs linked to milk intake.
Broken assumption
MicroRNAs in food survive digestion and enter human blood.
Why
The abstract does not give a cause. The results do not support the transfer of microRNAs from food into human blood.
Field: Psychology › Social PsychologyType: No effect foundArea: primate cognitionTarget: Action imitation in chimpanzeesStage: full study
Hypothesis
Untrained chimpanzees that watched a rewarded chimpanzee perform an action would copy it, even with no object or apparatus to learn from.
What failed
All but one chimpanzee failed to copy familiar actions. When tested with a new action, the one successful chimpanzee also failed to copy it, in several contexts.
Broken assumption
Chimpanzees naturally imitate the actions of other chimpanzees.
Why
Earlier positive studies may have shown chimpanzees learning how objects move rather than copying actions. Without that physical information, they did not imitate.
Field: Medicine › OncologyType: No effect foundArea: pediatric sarcomaTarget: STAT3 inhibitor LY5Stage: preclinical
Hypothesis
LY5, a new small molecule designed to block STAT3, would slow the growth and spread of osteosarcoma, Ewing's sarcoma and rhabdomyosarcoma.
What failed
LY5 killed sarcoma cells in culture and was well absorbed by mouth, but it did not slow tumour growth or prevent lung metastasis in mice. Knocking down STAT3 itself did not reproduce LY5's effects on the cells.
Broken assumption
A compound that blocks STAT3 activation in cells kills sarcoma cells because it blocks STAT3.
Why
LY5's effect on cells seems to come from targets other than STAT3, and it needed much higher concentrations to block metastasis than to block STAT3.
Field: Immunology and Microbiology › ImmunologyType: No effect foundArea: b cell immunologyTarget: ZBTB38 transcription factorStage: preclinical
Hypothesis
ZBTB38, a transcription factor highly expressed in germinal centre B cells and plasma cells, would be needed for normal antibody responses, as related family members are.
What failed
Mice lacking ZBTB38 in blood cells had normal germinal centre B cell numbers, normal antibody responses after immunisation, and normal memory B cell recall. Very few genes changed in B cells without ZBTB38.
Broken assumption
A transcription factor that is highly expressed in antibody-producing B cells is needed for antibody responses.
Why
The abstract gives no mechanism. ZBTB38 is not needed for these responses even though it is highly expressed in the cells that make them.
Field: Medicine › HepatologyType: No effect foundArea: hepatocellular carcinomaTarget: Transjugular intrahepatic portosystemic shunt (TIPS)Stage: full study
Hypothesis
A TIPS shunt changes blood flow in the liver, so it would interfere with imaging and delay the diagnosis of liver cancer.
What failed
Among 279 patients, the 37 with a TIPS had more scans and a longer time from first scan to diagnosis, but their LI-RADS imaging criteria, tumour stage at diagnosis, transplant rate and survival did not differ.
Broken assumption
A TIPS shunt makes liver cancer harder to diagnose on contrast imaging.
Why
The shunt lengthened surveillance but did not make tumours harder to classify on imaging or leave them at a later stage.
Field: Neuroscience › Cognitive NeuroscienceType: Couldn't replicate a published resultArea: psycholinguisticsTarget: Partner-specific memory associations in picture namingStage: replication
Hypothesis
Speakers would name a pictured object faster in front of a person they had earlier linked with that object's name, as Horton (2007, Experiment 1) found.
What failed
Three experiments designed to replicate and extend Horton's Experiment 1 did not reproduce the effect.
Broken assumption
Memory links between a conversation partner and a word speed up naming in that partner's presence.
Field: Neuroscience › Endocrine and Autonomic SystemsType: Couldn't replicate a published resultArea: obesity and glucose metabolismTarget: GPR21 receptorStage: preclinical
Hypothesis
Mice lacking the GPR21 receptor would resist diet-induced obesity and have better glucose tolerance, as an earlier knockout line showed.
What failed
A new GPR21 knockout line made with TALEN gene editing did not keep the better metabolic profile on a normal or high-fat diet.
Broken assumption
The metabolic effect of a knockout mouse comes from the gene that was deleted.
Why
The earlier knockout also disturbed expression of the neighbouring Rabgap1 gene, which may have caused the original effect.