,
15 tweets,
15 min read
Read on Twitter
@metapredict The GEO repository did not specify which groups were used for the ROC curves. I offered Dr Timmons to re-run my analysis if he sent me this information but he didn't and when the editor asked him to provide a code reproducing his results, he declined and threatened legal action. 
@metapredict Most people I told about this answered that it was nonsense as there
was no way such an action could be won. Eventually there was indeed no
action, but I don't think these threats are a nonsense.
was no way such an action could be won. Eventually there was indeed no
action, but I don't think these threats are a nonsense.
@metapredict The editor only asked for cosmetic changes in my correspondence, yet
it took more than two years to be published. Mostly waiting for Genome
Biology's legal counsel and editorial integrity departments to give
green lights.
it took more than two years to be published. Mostly waiting for Genome
Biology's legal counsel and editorial integrity departments to give
green lights.
@metapredict Professor @d_melzer and @lcpilling also reported that they eventually
had to give up on publishing their own critique, even after it was
accepted by Genome Biology. This was reported by the @bmj_latest.
bmj.com/content/362/bm…
had to give up on publishing their own critique, even after it was
accepted by Genome Biology. This was reported by the @bmj_latest.
bmj.com/content/362/bm…
@metapredict @d_melzer @lcpilling @bmj_latest When he was informed that my correspondence was about to be published
Dr Timmons emailed the editor, claiming that his signature stood out
from random samples by just removing some genes from the sampling
pool.
Dr Timmons emailed the editor, claiming that his signature stood out
from random samples by just removing some genes from the sampling
pool.
@metapredict @d_melzer @lcpilling @bmj_latest I tried myself, and found that this was not true. I informed Genome
Biology and suggested they had the experiment re-done independently.
They told me that a reviewer confirmed what I said.
Biology and suggested they had the experiment re-done independently.
They told me that a reviewer confirmed what I said.
@metapredict @d_melzer @lcpilling @bmj_latest Dr Timmons' email again came with legal threats if the
correspondence was to be published before his response was ready. Two
years after being offered a chance to respond, one year after stating
that he would not respond in Genome Biology and starting his threats.
correspondence was to be published before his response was ready. Two
years after being offered a chance to respond, one year after stating
that he would not respond in Genome Biology and starting his threats.
@metapredict @d_melzer @lcpilling @bmj_latest Genome Biology published anyway but added this note above the
correspondence. As expected, its claims are not supported by the
published response. The note has now been silently removed.
correspondence. As expected, its claims are not supported by the
published response. The note has now been silently removed.
@metapredict @d_melzer @lcpilling @bmj_latest Genome Biology chose to display above the correspondence a claim which
they knew was wrong, made by someone who was threatening a lawsuit,
without disclosing these threats or the fact that he had declined to
publish a peer-reviewed response over a two year period.
they knew was wrong, made by someone who was threatening a lawsuit,
without disclosing these threats or the fact that he had declined to
publish a peer-reviewed response over a two year period.
@metapredict @d_melzer @lcpilling @bmj_latest I regret to see that legal threats against Genome Biology are
effective. They should be disclosed by the journal, and should not
affect the editorial process to this extent.
effective. They should be disclosed by the journal, and should not
affect the editorial process to this extent.
@metapredict @d_melzer @lcpilling @bmj_latest The response still shows no evidence that the 150 HAGS probe-sets
stand out from randomly sampled sets of 150. Figure 1 shows the HAGS
standing out from a boxplot, but it was combined with 48 additional
predictive genes which were selected using Cohort 1.
stand out from randomly sampled sets of 150. Figure 1 shows the HAGS
standing out from a boxplot, but it was combined with 48 additional
predictive genes which were selected using Cohort 1.
@metapredict @d_melzer @lcpilling @bmj_latest These 48 were not added to the random 150s defining the boxplot. When
they are added, the HAGS does not stand out. I asked for these numbers
(150+48 vs 150) to be explicitly displayed on the figure but this was
fought off by Dr Timmons and eventually denied.
they are added, the HAGS does not stand out. I asked for these numbers
(150+48 vs 150) to be explicitly displayed on the figure but this was
fought off by Dr Timmons and eventually denied.
@metapredict @d_melzer @lcpilling @bmj_latest The response also claims that the HAGS stands out from random
samplings when looking across 4 muscle datasets. I had done the same
experiment, showing that this advantage disappeared when filtering out
poor quality probesets from the sampling pool.
samplings when looking across 4 muscle datasets. I had done the same
experiment, showing that this advantage disappeared when filtering out
poor quality probesets from the sampling pool.
@metapredict @d_melzer @lcpilling @bmj_latest I was not allowed to mention this experiment in my correspondence,
but made it publicly available upon its publication. I asked it to be
taken into account in the response, but this was also denied.
disq.us/p/1u97kte
but made it publicly available upon its publication. I asked it to be
taken into account in the response, but this was also denied.
disq.us/p/1u97kte
@metapredict Dr Timmons answered this, and blocked me to prevent me from answering. Screen shots prove no such thing, and here is the email I got from the editor in March 2017.
