#Medsky🧪 #Publichealth @nature.com
Some disease-predictors use dodgy data . Dozens of AI models designed to predict a person’s risk of #stroke or #diabetes are trained on dubious datasets. Researchers found that many such models used one of two open-access health datasets

“All models are wrong, but some are useful”
- George E.P. Box

Are we misled by myocardial infarction (MI) models?

Open-chest MI triggers systemic inflammation that can mask MI-specific immune responses

Minimally invasive MI tells a different story

Read our new preprint🧪
doi.org/10.64898/202...

Genetics may help explain why results from weight-loss jabs vary, say scientists Data on almost 28,000 patients suggests understanding gene variations could improve treatments for obesity

🧪 #BadJournalism
Headline: Genetics don’t explain variability in weight-loss jab effects

“The authors’ model suggests… genetics adding only a modest incremental contribution.”

Boring non-genetic factors such as sex, drug type, dose and duration most important

www.theguardian.com/science/2026...

If you’re using 10x Genomics (Flex v1) for scRNA-seq, this study is a must-read. 🧪

It uncovers a key artefact:
• Probe barcodes themselves introduce bias
• Some barcodes systematically alter gene expression signals
• Apparent DEGs may be technical, not biological

www.biorxiv.org/content/10.6...

major issues with antibodies used in #neuroscience research
www.thetransmitter.org/open-neurosc...
#reproducibility

🔬 Are you using the right antibody for your experiments?

Only Good Antibodies helps researchers identify specific antibodies

🧪 1496 antibodies against 147 proteins tested so far.

Check now
oga-website.onrender.com

OGA is powered by YCharOS

🧪

A step forward for cardio-immunology!

Our recent preprint shows:

Standard MI models distort neutrophil responses through surgical inflammation.

A minimally invasive model better mirrors patients and reveals immature neutrophils driving post-MI damage.

Here is link www.biorxiv.org/content/10.6...

🧪

How specific are therapeutic monoclonal antibodies?

In a study of 174 FDA-approved antibodies, 28% showed off-target binding.

Surprisingly, even the famous pembrolizumab (anti–PD-1) binds another protein (TDGF1).

www.sciencedirect.com/science/arti...

@biorxivpreprint.bsky.social A Genetic Tool for Specific Tracking of Mature Neutrophils
www.biorxiv.org/content/10.6...

On the balance of knowledge - Nature Reviews Immunology Ruslan Medzhitov shares his thoughts on the balance between generating data and developing theories in immunology, with a focus on exploring the rules that govern complex systems.

I'd like to see this problem analysed at the single cell multi-omics level
'Advances in research technolo- gies are crucial for the scientific enterprise but have an unintended consequence: when pow- erful techniques become widely available, they tend to be applied without a good question in mind.'

AI Doesn’t Reduce Work—It Intensifies It One of the promises of AI is that it can reduce workloads so employees can focus more on higher-value and more engaging tasks. But according to new research, AI tools don’t reduce work, they consisten...

Engineers increasingly found themselves coaching colleagues who were “vibe-coding” and finishing partially complete pull requests. This oversight often surfaced informally—in Slack threads or quick desk-side consultations—adding to engineers’ workloads.'
hbr.org/2026/02/ai-d...

🧪

Do we study the right biology after myocardial infarction (MI)?
• Surgery in standard mouse model disrupts bone marrow & masks MI-specific immune responses
• Minimally invasive model better reflects patient immune responses & reveals immature neutrophils worsen heart function
tinyurl.com/MImodels

Scientific paper mill: the model flagged 10% of the research papers in cancer research as problematic 😱

www.bmj.com/content/392/...

A study shows the consequences of poor Antibody Validation

- Out of 614 antibodies tested, 97 failed
- These 97 non-specific antibodies (43 proteins) were used in 640 publications
- 9443 mice and 5015 human samples wasted

Waste of animals, human samples, time, and trust

tinyurl.com/OGA-Ab

The antibodies don’t work! The race to rid labs of molecules that ruin experiments Poorly performing antibodies have plagued biomedical sciences for decades. Several fresh initiatives hope to change this.

unfortunately, several scientists experienced the same situation. Worth to read this article www.nature.com/articles/d41...

The statement is maybe strong, but how often Ab specificity is tested?
My experience: A study claimed new biology using non-validated Abs. After we showed the Abs were unsuitable, the supplier discontinued them. Meanwhile, study has been cited 150+ times. This is how artefacts can become “facts".

For instance ab263899 that recognises human NLRP3 specifically was cited 737 times, according to CiteAb (see image)

CiteAb - Reagent Search & Life Science Data CiteAb - Reagent Search & Life Science Data

Good question.
I would suggest to use CiteAb website. They catalogue all commercial antibodies and report how many times each antibody has been cited. In addition, they provide information on antibody validation.
www.citeab.com

Highly recommend using CiteAb and check for genetic (KO/KD) validation too.

In all seriousness,use CiteAb and Antibodypedia, look for KO/KD validated antibodies, best - recombinant, second best - monoclonal, and if possible do your own KO/KD controls, or induction by ligand/small molecule,if not - compare hyperexpressing with low expressing cells (from depmap/protein atlas)

VA-LI-DA-TE your abs! My experience (and some spooky examples in the thread): bsky.app/profile/gree...

That would be an interesting study to compare both models.

🧪

A minimally invasive thrombotic model to study stroke in awake mice - Nature Communications Rodent stroke models are crucial but often invasive and limited in translational relevance. Here, the authors show that magnetic particle delivery creates a reproducible, minimally invasive thrombotic...

A new magnetic particle mouse model of stroke

Importantly, no craniotomy is required, avoiding a major confounder that causes massive surgery-induced inflammation.

Read the article "A minimally invasive thrombotic model to study stroke in awake mice"
www.nature.com/articles/s41...

Nature Medicine to investigate study that found cancer treatment is better in morning Nature Medicine starts an investigation into inconsistencies in a study that found it was better to have immunotherapy in the morning.

www.statnews.com/2026/02/20/c...
i hope they can get the data from the authors. I was told by the editor that data weren’t available because of patient confidentiality- so many authors use that excuse , it makes data sharing policies worthless

This is weird. On @pubpeer.com, a reader noticed some seemingly irrelevant references in a Scientific Reports paper pubpeer.com/publications... The paper has now been modified, with many references removed (from 45 down to 25!) but it contains no acknowledgment of the change. It does have EoC.

🧪