The TDR Targets Database

Expression evidence was collected from several microarray experiments and combined in a simplified scheme. Genes were grouped in 5 categories depending on how much upregulation they showed in the selected life cycle stage: 0-20%, 20-40%, 40-60%, 60-80% and 80-100%. With 0-20% representing the lower 20% of genes, showing less upregulation in the corresponding life cycle stage; and 80-100% representing the top 20% of genes (i.e. those showing a higher upregulation).

More detailed queries (fine grained) on gene expression can be performed with the original datasets at their respective sources (for example PlasmoDB).

Putative antigenic peptides were predicted using the method of Kolaskar and Tongaonkar, as implemented in EMBOSS (antigenic). Each predicted epitope has an associated score based on the physicochemical properties of the aminoacid residues. A cumulative antigenicity score was calculated as the sum of scores of all predicted epitopes in the protein. A normalized antigenicity index was calculated as the ratio of this cumulative score over the protein length. Finally, for each antigenicity index value we calculated the percent of proteins for each genome that fall below it. Based on this, we implemented the antigenicity index percentile query above. Thus a search for antigenicity index percentile > 80 will fetch all proteins with an antigenicity index such that 80% of the observations fall below it (i.e. the proteins with the top 20% index values).
Reference: A semi-empirical method for prediction of antigenic determinants on protein antigens. Kolaskar AS and Tongaonkar PC. FEBS Lett 276: 172. PubMed

Note that you cannot use both the simplified form above and the advanced form below at the same time. This will produce unpredictable results! You have been warned!

To annotate phenotypes and/or validation experiments, we are using a phenotype syntax composed of terms derived from a number of controlled vocabularies (i.e. ontologies, e.g. GO, PATO, ECO). The syntax we are using is similar in spirit to the Pheno-syntax format described by Chris Mungall, although we have not attempted to formally comply with the pheno-syntax grammar.

The syntax consists of a collection of tag-values. Tags are i) the phenotypic quality or attribute (Q), ii) the entity bearing the phenotype (B), iii) an anatomy term that describes where the phenotype occurs (A) — not relevant for unicellular organisms, and iv) a term describing the timing during which the phenotype occurs (T) – usually a developmental stage.

As an example, the way to describe "slow growth in bloodstream forms" using this syntax is:

Phenotypic Quality (Q) = slow
Affected entity (B) = growth
During (T) = bloodstream form
Where (A) = whole organism

In this case the affected entity is growth and the way in which the growth is affected is described by the Q term (slow in this case), and this phenotype occurs in the whole organism.

The form below gives you access to search for targets using any of the terms we have used to annotate them. Do note that only a few combinations of Q, B, A, and T terms are represented in the db, even though you can search for all of them!

Phenotypes:

Validation:

Phenotypes/Validation:

The druggability index query lets you restrict your search based on a druggability measure. Where the information about druggability is not available from the pathogen protein itself, it is derived from the closest druggable homolog.

The druggability index (Dindex) is a composite score consisting of a weighted normalised sum, where each of the different druggability prediction methods are given different weights depending on their relative contribution to prediction. The Dindex values range from 0 to 1, where a larger index score means a more likely to be druggable target.

By doing sequence similarity searches (using BLAST) against a database of known targets derived from the latest Inpharmatica literature SAR database (Starlite) a large number of proteins in TDR priority species could be linked to a known druggable target with at least 1 small molecule compound with a binding affinity less than 10um.

Do note that the Dindex has been calculated originally for known druggable targets in Starlite and has been mapped to pathogen targets based only on BLAST-based similarity.

The compound desirability query lets you restrict your search to targets based on the chemical quality of the compounds associated with the target (or with its closest druggable homolog).

The compound desirability index (Pfizer) was calculated for the compounds associated with each of the closest druggable targets (as measured by BLAST similarity, see above). The compound desirability index is a fitness value that summarises the average `chemical quality' of each target.
The compound desirability value links direct to actual compounds (which could the basis for composing target-specific screening subsets of compounds), but these have not been disclosed and are not available for searching and/or display.

The desirability function is based on Harrington's desirability index where the target function is based on the molecular properties distribution of oral, small molecule drugs. The function also contains penalty functions for acticity, promiscuity and structural alerts (risk and reactive groups in compounds).

Associated compounds: associations between genes and chemical compounds are derived from a number of sources. In each case the association between a gene and a compound has been done by different methods, and thus the reliability and/or the relevance of the association is different for each source.

The sources are: i) manual curation (more reliable, more relevant); ii) transitive (semi-automatic) association of compounds to gene A based on its similarity to a gene B for which there are known compounds (DrugBank); iii) finally associations between compounds and enzymes have been established based on co-occurrence of EC numbers and compound identifiers (CAS registry IDs) in PubMed abstracts (indexed for MEDLINE). In this last case because the link between the gene and the compound is established only due to co-occurrence, there is no guarantee of any functional association between the targets and the associated compounds. In other words, there is no guarantee that genes annotated with an EC number will interact, bind to, or be inhibited by the compounds that are mentioned in the corresponding papers. Also, the compounds obtained in this way might not be the same as those for which the desirability index has been calculated. You have been warned.

Assayability is currently evaluated as positive if the target is an enzyme, and if the enzyme has an assay described in the list of enzyme assays available from Sigma-Aldrich. The availability of the protein in recombinant, soluble form is evaluated as positive based on the data provided by the Consortium for Structural Genomics of Pathogenic Protozoa. This is based on their ability to express the protein in an heterologous system (usually E. coli) in soluble form. You can check the status of their progress here.