The Supplementary Materials (SM) for the Science article provides references to a number of different journal-articles that outline various parts of the methodology.
The SM can be found here
Extending the time-series to include Landsat-5 (and potentially Landsat-8 to make the methodology something that can be rerun "easily") data will be a challenging task, and will require extensive testing of the image normalization. The image normalization may be made even more difficult, due to the lack of coinciding MODIS coverages. Instead, you may have to apply a different normalization approach, such as a combination of COST - article and MAD - article as outlined in this article.
All-in-all, it is not a simple task, and I wish you good luck with it.
The most relevant articles from the SM are:
P. Potapov, S. A. Turubanova, M. C. Hansen, B. Adusei, M. Broich, A. Altstatt, L. Mane, C. O. Justice, Quantifying forest cover loss in Democratic Republic of the Congo, 2000-
2010. Remote Sens. Environ. 122, 106–116 (2012). Article
M. Broich, M. C. Hansen, P. Potapov, B. Adusei, E. Lindquist, S. V. Stehman, Time-series analysis of multi-resolution optical imagery for quantifying forest cover loss in Sumatra
and Kalimantan, Indonesia. Int. J. Appl. Earth Obs 13, 277–291 (2011).
M. Hansen, A. Egorov, D. P. Roy, P. Potapov, J. Ju, S. Turubanova, I. Kommareddy, T. R. Loveland, Continuous fields of land cover for the conterminous United States using Landsat data: First results from the Web-Enabled Landsat Data (WELD) project. Remote
Sens. Letters 2, 279–288 (2011). Article
M. Hansen, R. S. DeFries, J. R. G. Townshend, M. Carroll, C. Dimiceli, R. A. Sohlberg, Global percent tree cover at a spatial resolution of 500 meters: First results of the MODIS vegetation continuous fields algorithm. Earth Interact. 7, 1–15 (2003). Article
L. Breiman, J. Friedman, R. Olsen, C. Stone, Classification and Regression Trees Wadsworth and Brooks/Cole, Monterey, CA, (1984).