Customization scenarios for de-identification of clinical notes.

BACKGROUND: Automated machine-learning systems are able to de-identify electronic medical records, including free-text clinical notes. Use of such systems would greatly boost the amount of data available to researchers, yet their deployment has been limited due to uncertainty about their performance when applied to new datasets. OBJECTIVE: We present practical options for clinical note de-identification, assessing performance of machine learning systems ranging from off-the-shelf to fully customized. METHODS: We implement a state-of-the-art machine learning de-identification system, training and testing on pairs of datasets that match the deployment scenarios. We use clinical notes from two i2b2 competition corpora, the Physionet Gold Standard corpus, and parts of the MIMIC-III dataset. RESULTS: Fully customized systems remove 97-99% of personally identifying information. Performance of off-the-shelf systems varies by dataset, with performance mostly above 90%. Providing a small labeled dataset or large unlabeled dataset allows for fine-tuning that improves performance over off-the-shelf systems. CONCLUSION: Health organizations should be aware of the levels of customization available when selecting a de-identification deployment solution, in order to choose the one that best matches their resources and target performance level.

Tools and strategies for physiological genomics: the Rat Genome Database.

The broad goal of physiological genomics research is to link genes to their functions using appropriate experimental and computational techniques. Modern genomics experiments enable the generation of vast quantities of data, and interpretation of this data requires the integration of information derived from many diverse sources. Computational biology and bioinformatics offer the ability to manage and channel this information torrent. The Rat Genome Database (RGD; http://rgd.mcw.edu) has developed computational tools and strategies specifically supporting the goal of linking genes to their functional roles in rat and, using comparative genomics, to human and mouse. We present an overview of the database with a focus on these unique computational tools and describe strategies for the use of these resources in the area of physiological genomics.

ZoomQuant: an application for the quantitation of stable isotope labeled peptides.

The main goal of comparative proteomics is the quantitation of the differences in abundance of many proteins between two different biological samples in a single experiment. By differentially labeling the peptides from the two samples and combining them in a single analysis, relative ratios of protein abundance can be accurately determined. Protease catalyzed (18)O exchange is a simple method to differentially label peptides, but the lack of robust software tools to analyze the data from mass spectra of (18)O labeled peptides generated by common ion trap mass spectrometers has been a limitation. ZoomQuant is a stand-alone computational tool that analyzes the mass spectra of (18)O labeled peptides from ion trap instruments and determines relative abundance ratios between two samples. Starting with a filtered list of candidate peptides that have been successfully identified by Sequest, ZoomQuant analyzes the isotopic forms of the peptides using high-resolution zoom scan spectrum data. The theoretical isotope distribution is determined from the peptide sequence and is used to deconvolute the peak areas associated with the unlabeled, partially labeled, and fully labeled species. The ratio between the labeled and unlabeled peptides is then calculated using several different methods. ZoomQuant's graphical user interface allows the user to view and adjust the parameters for peak calling and quantitation and select which peptides should contribute to the overall abundance ratio calculation. Finally, ZoomQuant generates a summary report of the relative abundance of the peptides identified in the two samples.

Simultaneous quantification and identification using 18O labeling with an ion trap mass spectrometer and the analysis software application "ZoomQuant".

Stable isotope labeling with (18)O is a promising technique for obtaining both qualitative and quantitative information from a single differential protein expression experiment. The small 4 Da mass shift produced by incorporation of two molecules of (18)O, and the lack of available methods for automated quantification of large data sets has limited the use of this approach with electrospray ionization-ion trap (ESI-IT) mass spectrometers. In this paper, we describe a method of acquiring ESI-IT mass spectrometric data that provides accurate calculation of relative ratios of peptides that have been differentially labeled using(18)O. The method utilizes zoom scans to provide high resolution data. This allows for accurate calculation of (18)O/(16)O ratios for peptides even when as much as 50% of a (18)O labeled peptide is present as the singly labeled species. The use of zoom scan data also provides sufficient resolution for calculating accurate ratios for peptides of +3 and lower charge states. Sequence coverage is comparable to that obtained with data acquisition modes that use only MS and MS/MS scans. We have employed a newly developed analysis software tool, ZoomQuant, which allows for the automated analysis of large data sets. We show that the combination of zoom scan data acquisition and analysis using ZoomQuant provides calculation of isotopic ratios accurate to approximately 21%. This compares well with data produced from (18)O labeling experiments using time of flight (TOF) and Fourier transform-ion cyclotron resonance (FT-ICR) MS instruments.