----- Original Message -----
From: Hamilton, James
To: Tianbo Jin ; DreesB@umkc.edu ; papasianc@umkc.edu
Cc: Yang, Runqing ; Deng, Hong-Wen
Sent: Tuesday, December 09, 2008 11:50 AM
Subject: RE: Whole genome association papers in Dr. Hong-Wen Deng's group
I certainly agree! I don't personally have detailed knowledge of the data or methods presented in the paper. If I receive a paper to review from Dr. Deng's lab, I have assumed that at a minimum the science involved is good. I will not accept that we place "intentionally untrue results" in papers originating from my Department. If that means we need to do more detailed experiments or collect a larger sample to make the results valid, then we must do that before the paper is submitted. The push to get papers published is NOT a justification for bad science.
I would hope the senior researchers who are intimately involved in the oversight of these papers would rigidly enforce the basic requirement of good science before publication. I cannot tell you how distressing I find this situation. I am sure that Drs. Drees and Papapasian will have a similar response.
Dr. Deng do you have any suggestions as to how this concern can be corrected?
Jim Hamilton
--------------------------------------------------------------------------------
From: jint@umkc.edu
Sent: Tue 12/9/2008 11:31 AM
To: DreesB@umkc.edu; Hamilton, James; papasianc@umkc.edu
Cc: Yang, Runqing; Deng, Hong-Wen
Subject: Whole genome association papers in Dr. Hong-Wen Deng's group
Dear Professors,
You must be very careful when you revise the manual scripts for our group. I did data analysis from the beginning. There are some intentionally untrue results in our manual scripts. Below are some examples of untrue results:
1. For whole genome association (WGA) papers, the call rates listed in the MSes were not true. Eleven of one thousand Caucasians were identified with call rates lower than 93% (Table below). We didn’t do any replication for the samples. Dr. Deng still suggested use one thousand subjects to perform the GWA studies (Attachment).
2. Population stratification can affect the results of WGA studies. Ideal subjects should come from the same population when they were used to perform WGA analysis. I detected the population stratification in our one thousand subjects. I could not detect the stratification when I used 200 unrelated Probes. I did detect stratification in our subjects when I used 2,000 and 5,000 unrelated probes. They only adopted the results gotten from 200 unrelated probes. Actually, the results we have gotten from 2,000 or 5,000 unrelated probes should be more accurate than that from 200 unrelated probes. The WGA results would have no value if we admitted that there was stratification in our sample. So the results of population stratification in our MSes were not true.
3. We performed the WGA and Bivariate analyses using the wrong raw data. We have prepared so many MSes ready to be published. We got significant results for every phenotype in GWA studies. I didn’t think that was reasonable and true. For the same phenotype, we can get many different results if was analyzed by different people in our group. There was no quality control. The results were not reliable.
4. We are doing the whole genome pathway analysis right now. The method has been identified wrong by Dr. Runqing Yang, a visiting scholar in our group. We still perform the analysis and have distributed the related papers to our group.
5. I also found some male subjects had the age at menarche (AAM) data in our original database.
Above are just the examples of our untrue data and results. I don’t think that we can do science like this.
Table 1. Samples with call rates lower than 93%
Sample ID
Gender
Call rate
498HD0051
male
0.471259
498HD0029
female
0.510834
498HD0668
female
0.518536
498HD0130
female
0.519165
498HD0376
male
0.784372
498HD0036
male
0.832965
498HD0409
female
0.867669
498HD0062
female
0.869009
498HD0354
female
0.898031
498HD0056
male
0.921724
498HD0245
female
0.929757
Best Wishes,