After postdoctoral research at the University of Southern California from 1991-1992 and at Sandia National Laboratories in Albuquerque from 1992-1993, she took a faculty position at the University of Pennsylvania, where she remained until moving in 1999 to the University of Texas. In 2014, Warnow joined the faculty of the University of Illinois in Urbana-Champaign, where she is the Founder Professor of Engineering and Associate Head of the Department of Computer Science. Warnow has courtesy appointments in the Departments of Animal Biology, Bioengineering, Electrical and Computer Engineering, Entomology, Mathematics, Plant Biology, and Statistics. In 1995, research by Warnow, Donald Ringe, and Ann Taylor at the University of Pennsylvania based on perfect phylogeny computations provided a comprehensive theory for the timing of the early subdivisions in the Indo-European languages. Their computations lent support to the Indo-Hittite hypothesis according to which the first of these subdivisions to separate from the rest of the Indo-European languages were the Anatolian languages. Their results also support the Graeco-Armenian hypothesis, according to which the Armenian language and Greek language form a subfamily of Indo-European. They fit the Germanic languages into the evolutionary tree of Indo-European languages, previously considered problematic, by hypothesizing that the Proto-Germanic language was closely related to the Balto-Slavic languages but then became modified by westward migrations of the Germanic tribes which led them into contact with Italic and Celtic speakers. This perfect phylogeny approach was later extended by Warnow and colleagues to allow for undetected borrowing between languages, so that language evolution is modelled with a network rather than a tree. In 2009, Warnow and her colleagues released their SATé software for co-estimating biological multiple sequence alignments and evolutionary trees. Their software is based less strongly on firm mathematical principles than some previous co-estimation methods, but is significantly faster, allowing the fast construction of highly accurate trees and alignments for thousands of species. In comparison, the slow performance of previous methods limited them to only comparing dozens of species at a time. Her work from 2014-2018 has focused on three topics: scaling multiple sequence alignments to ultra-large datasets, species tree estimation using multiple genes, and metagenomics. Her major contributions in these topics include the PASTA method for co-estimation of alignments and trees, which improves on SATé, and can produce highly accurate alignments with up to 1,000,000 sequences. She has also developed the ASTRAL method for species tree estimation, which is a statistically consistent method for constructing species trees in the presence of incomplete lineage sorting.
