Catabolism

Transwell invasion assay was performed as described previously^{4 (link)}. In brief, cells were loaded onto the upper well of the Transwell chamber with 8 µm ϕ pore membrane (Coster), precoated with Matrigel on an upper side of the chamber. The lower well was filled with 600 µl of DMEM containing 10% FBS. After incubation for 24 hr, cells invaded to lower surface of the membrane were counted. For ECM degradation assay, glass coverslips were coated with gelatin conjugated with either Alexa Fluor 594 (Invitrogen) (Alexa-gelatin) or fluorescein (Invitrogen) (FL-gelatin) as described^{65 (link)}. Transfected cells were trypsinized, replated on these glass coverslips, and cultured for 6 hr. After fixation, cells were fixed and stained with phalloidin. Number of invadopodia, identified as F-actin dots in the areas of degraded gelatin, was quantified by using the ImageJ particle analysis tool.

Protein sequences were collected from the Swiss-Prot database at http://www.ebi.ac.uk/swissprot/. The detailed procedures are basically
the same as those elaborated in [13] (link); the only differences are as follows.
(1) To get the updated benchmark dataset, instead of version 49.3 of the
Swiss-Prot database, the version 55.3 released on 29-Apr-2008 was adopted.
(2) In order to make the new predictor also able to deal with proteins
having two or more location sites, the multiplex proteins are no longer excluded in this
study. Actually, according to a statistical analysis on the current database, about
8% of plant proteins were found located in more than one location.
After strictly following the aforementioned procedures, we finally obtained a benchmark
dataset containing 978 different protein sequences, which are distributed
among 12 subcellular locations (Fig. 1); i.e., where represents the subset for the subcellular location of cell membrane, for cell wall, for chloroplast, and so forth; while represents the symbol for “union” in the set
theory. A breakdown of the 978 plant proteins in the benchmark dataset according to their 12 location sites is given in Table 1. To avoid redundancy and homology bias, none of the proteins in has pairwise sequence identity to any other in a same subset. The
corresponding accession numbers and protein sequences are given in Table S1.
Since some proteins in may occur in two or more locations, it is instructive to introduce the
concept of “locative protein” [23] (link), as briefed as follows. A
protein coexisting at two different location sites will be counted as 2 locative
proteins even though the two are with completely the same sequence; if
coexisting at three sites, 3 locative proteins; and so forth. Thus, it follows where is the number of total locative proteins, the number of total different protein sequences, the number of proteins with one location, the number of proteins with two locations, and so forth;
while is the number of total subcellular location sites concerned (for the
current case, as shown in Fig. 1).
For the current 978 different protein sequences, 904 occur in one subcellular location,
71 in two locations, 3 in three locations, and none in four or more locations.
Substituting these data into Eq.2, we have which is fully consistent with the figures in Table 1 and the data in Table S1.
To develop a powerful method for predicting protein subcellular localization, it is very
important to formulate the sample of a protein in terms of the core features that are
intrinsically correlated with its localization in a cell. To realize this, the strategy
by integrating the GO representation and PseAAC representation was adopted in the
original Plant-PLoc [13] (link). In this study, the essence of such a strategy will be
still kept. However, in order to overcome the four shortcomings as mentioned in Introduction for Plant-PLoc [13] (link), a completely different
combination approach has been developed, as described below.