TagSmart: analysis and visualization for yeast mutant fitness data measured by tag microarrays

Chulyun Kim, Sangkyum Kim, Russell Dorer, Dan Xie, Jiawei Han, Sheng ZhongBMC Bioinformatics 2007 · DOI

Abstract

TagSmart is a web-based analysis and visualization tool designed to handle complex fitness data from yeast mutant libraries. By implementing a robust statistical framework and a novel "matched permutation" strategy, it allows researchers to accurately identify gene-environment interactions across multiple time points and drug dosages with a remarkably high validation rate.

Introduction

In functional genomics, the yeast Saccharomyces cerevisiae is a primary model organism. Researchers have developed nearly complete collections of "knockout" mutants, where specific genes are replaced with unique DNA barcodes known as "tags" (uptags and downtags). By growing these mutants in a single pool and using microarrays to quantify tag abundance, scientists can measure the fitness of thousands of mutants simultaneously.

However, analyzing this data presents significant challenges. Experiments often involve complex designs, such as measuring fitness over multiple generations or testing drugs at various concentrations. Existing software often lacks user-friendly interfaces, is limited to simple experimental setups, or fails to account for the inherent noise in microarray data, such as "bad tags" that do not hybridize correctly.

The TagSmart Architecture

TagSmart addresses these challenges through a three-module architecture:

  1. Data Preprocessing: This module integrates raw microarray files (CEL or TXT) with chip descriptions and experimental metadata. It allows users to "mask" known bad tags that might otherwise skew results.
  2. Statistical Computation: Instead of simple averaging, TagSmart utilizes a modified T-statistic to compare concentration changes between treated and control conditions. To determine significance, it employs a matched permutation strategy. This approach generates a background distribution by shuffling data points within the same time, tag, and probe constraints, allowing for accurate calculation of q-values (False Discovery Rate).
  3. Visualization: The tool generates interactive heatmaps where rows represent mutants (and their four tag-probe pairs) and columns represent experimental conditions.
The TagSmart interface integrates statistical results with an interactive heatmap, allowing users to explore gene fitness across generations.

Key Insights: Probe-Level Analysis and Filtering

A critical feature of TagSmart is its formalization of the "bad tag" filter. Mutations in synthetic DNA tags during library construction or cross-hybridization on the array can create outliers. TagSmart provides a systematic way to identify and exclude these, significantly boosting sensitivity. Furthermore, the tool retains individual probe signals (Perfect Match vs. complementary Perfect Match) throughout preprocessing, preserving the granular data necessary for robust statistical analysis.

Experimental Validation and Results

The authors validated TagSmart using two distinct datasets:

  • Titration Mixtures: A controlled "ground truth" dataset where mutant strain concentrations were manually adjusted. TagSmart achieved recall rates of 0.7 to 0.9 at a precision level of 0.4.
  • Cincreasin Drug Study: A real-world study where heterozygous deletion strains were treated with Cincreasin, a spindle checkpoint inhibitor.

In the Cincreasin study, TagSmart identified 53 sensitive mutant strains. Subsequent individual "patch tests" confirmed 52 out of 53 (98%) as truly sensitive. The tool successfully identified Mps1 as the primary target of Cincreasin and uncovered a suite of other sensitive mutants suggesting a chained molecular response to spindle checkpoint inhibition.

Interactive tooltips in the heatmap provide exact fitness values for specific data points, facilitating granular data exploration.

Conclusion

TagSmart provides a powerful, user-friendly platform for yeast researchers to analyze complex fitness data. While currently optimized for S. cerevisiae tag arrays, future work aims to integrate other "omics" data and adapt the tool for newer barcoding technologies like Bar-seq.

Summary updated

BibTeX
@article{kim2007tagsmart,
  author = {Chulyun Kim and Sangkyum Kim and Russell Dorer and Dan Xie and Jiawei Han and Sheng Zhong},
  title = {{TagSmart: analysis and visualization for yeast mutant fitness data measured by tag microarrays}},
  journal = {BMC Bioinformatics},
  year = {2007},
  doi = {10.1186/1471-2105-8-128},
}
Figure