Proteomics - Summary of breakout discussion 17 Jan 2005

Participants: Billheimer (scribe), Chen, Koyama, M. Li, Shyr

1. What is the best way to collaborate with basic scientists in this area?

Problems in proteomics almost always require scientists/physicians from three areas of expertise: clinical (disease) specialists, measurement technology specialists, and quantitative specialists. Successful collaboration requires significant interaction and contribution from each group.

2. What is the best approach to determining the best approach for analyzing mass spec data?

We believe that comparative analysis of "known" standard data sets (e.g., uniformity data, spike-in data, numerical spike-in) currently provide the best approach for quantitative methods evaluation. Variation characteristics for MALDI-TOF protein spectra are extremely complex. Current simulation approaches do not reflect the noise characteristics of real spectra. (see question 3 below)

We currently have two standard data sets (100 spectra of uniform liver homogenate; 40 spectra from a protein spike-in experiment). Further, we believe that personnel in the Vanderbilt Mass Spectrometry Research Center will create new data sets per our request to evaluate data normalization and analysis methods.

3. Should a simulation model be developed for testing various methods?

Biostatisticians at MDACC are already doing this. We believe their current model's representation of variation is inadequate for rigorous methods testing. The VICC has invited Dr. Kevin Coombs to visit Vanderbilt in the near future. We will continue discussion at that time.

4. Should we try to better coordinate mass spec quantitative methodology development

• within Biostatistics
• across the University

We are working to establish a common groundwork among researchers within the Biostatistics Department. This includes verification of data pre-processing procedures, and gaining consensus on analysis methods for "standard" problems (e.g., class comparison). In addition we are establishing links to multiple collaborators with MALDI-TOF MS expertise.

There are many other quantitative scientists at Vanderbilt involved in proteometrics research. These include personnel in biochemistry, mathematics, electrical engineering, and bioinformatics. We believe Biostatistics does not have sufficient resources (FTE's) to be involved in these many disparate research efforts.

5. Are proteomic methodology education opportunities in existing journal clubs?

There are existing journal clubs discussing proteomic methodologies (e.g., the Proteomics Core lab journal club). However, these journal clubs focus on emerging measurement technologies - not on quantitative problems. (Indeed, they typically avoid quantitative issues.)

We propose two approaches to foster educational and collaborative opportunities for quantitative proteomics. These are

- create and maintain a Biostatistics TWiki page with links to quantitative proteomics papers

- begin a monthly lunch meeting for biostatisticians, biochemists, and others interested in quantitative aspects of proteomics (we anticipate 10 - 12 regular participants)

6. There are many other proteomic measurement methods in use/underdevelopment at Vanderbilt. How can we prepare to support these other technologies?

The first five question above all are related to proteomics via MALDI-TOf MS protein profiling. There are dozens of other emerging methods in proteomics including 2-D gel electrophoresis, 2-D DIGE, lipidomics, tandem MS-MS, structural NMR, and antibody-based molecular recognition assays. We should focus recruiting for a statistics/biostatistics candidate with (some) background and interest in this area.