Skip to main content

mastodon - Link to source

Preprint from Salzberg team questioning a 2020 Nature paper from Rob Knight 😮

"the raw read counts were vastly over-estimated for nearly every bacterial species, often by a factor of 1000 or more."

"Our conclusion after re-analysis is that the near-perfect association between microbes and cancer types reported in the study is, simply put, a fiction."

Major data analysis errors invalidate cancer microbiome findings

biorxiv.org/content/10.1101/20…

#microbiome #genomics #research #science


Major data analysis errors invalidate cancer microbiome findings

We re-analyzed the data from a recent large-scale study that reported strong correlations between microbial organisms and 33 different cancer types, and that created machine learning predictors with near-perfect accuracy at distinguishing among cance…
bioRxiv
#science #Research #genomics #microbiome
in reply to Feargal Ryan

Feargal Ryan
mastodon - Link to source
Feargal Ryan
• •

The authors of the original paper have also published a rebuttal on their github. THE DRAMA!

github.com/gregpoore/tcga_rebu…


GitHub - gregpoore/tcga_rebuttal: Re-analysis of data provided by Gihawi et al. 2023 bioRxiv

Re-analysis of data provided by Gihawi et al. 2023 bioRxiv - GitHub - gregpoore/tcga_rebuttal: Re-analysis of data provided by Gihawi et al. 2023 bioRxiv
GitHub
in reply to Feargal Ryan

Alex Crits-Christoph
mastodon - Link to source
Alex Crits-Christoph
• •
oh man. There's a lot to chew on in this but just read the section "Normalization of the reads erroneously created a distinct signature of each cancer" in the preprint. Many of the most important features in the classifier had 0 reads in all samples. That is reallyyy not good. The Github counter-rebuttal seems to be saying "But there's still a signal!" ignoring the huge flaws in the original...
This entry was edited (2 years ago)
in reply to Alex Crits-Christoph

Feargal Ryan
mastodon - Link to source
Feargal Ryan
• •
@alexcc I’ll be interested to see how the peer review plays out. But boy oh boy if what this preprint is saying is accurate these are some rookie mistakes to make. Like first year student with zero supervision type stuff...
@Alex Crits-Christoph
in reply to Feargal Ryan

Feargal Ryan
mastodon - Link to source
Feargal Ryan
• •
@alexcc “The models included species that had never been reported in humans, and that were associated only with extreme environments, ocean-dwelling species, plants, or other non- human environments.” - I’ve called this out when reviewing manuscripts on multiple occasions! It’s so basic
@Alex Crits-Christoph
in reply to Feargal Ryan

Frank Aylward
mastodon - Link to source
Frank Aylward
• •
@alexcc great example of why carefully preprocessing and validating the inputs for these models is so important.
@Alex Crits-Christoph

Lo, thar be cookies on this site to keep track of your login. By clicking 'okay', you are CONSENTING to this.

⇧