# aRrayLasso: a network-based approach to microarray interconversion

## TL;DR

aRrayLasso is a new method that improves the conversion of microarray data between different platforms using a statistical model.

## Contribution

aRrayLasso introduces a Lasso-penalized model for accurate microarray interconversion without relying on manufacturer annotations.

## Key findings

- aRrayLasso achieves high fidelity in predicting expression levels comparable to technical replicates.
- The method enables integration of datasets from different microarray platforms using open-source R functions.
- It outperforms existing methods that rely on direct probe alignment or incomplete annotations.

## Abstract

Summary: Robust conversion between microarray platforms is needed to leverage the wide variety of microarray expression studies that have been conducted to date. Currently available conversion methods rely on manufacturer annotations, which are often incomplete, or on direct alignment of probes from different platforms, which often fail to yield acceptable genewise correlation. Here, we describe aRrayLasso, which uses the Lasso-penalized generalized linear model to model the relationships between individual probes in different probe sets. We have implemented aRrayLasso in a set of five open-source R functions that allow the user to acquire data from public sources such as Gene Expression Omnibus, train a set of Lasso models on that data and directly map one microarray platform to another. aRrayLasso significantly predicts expression levels with similar fidelity to technical replicates of the same RNA pool, demonstrating its utility in the integration of datasets from different platforms.

Availability and implementation: All functions are available, along with descriptions, at https://github.com/adam-sam-brown/aRrayLasso.

Contact:
chirag_patel@hms.harvard.edu

Supplementary information:
Supplementary data are available at Bioinformatics online.