Related work

While there are a number of free and commercial tools that support one or more phases of HTS (e.g., Spotfire), in the following we focus our review on visualization tools that specifically address hit selection, exploration and expansion and more specifically the understanding of structure-activity relationships.

SARANEA [4] is a visualization tool to support structure-activity relationship and selectivity analysis. It is based on a network graph visualization. The graph is built by connecting molecules with an edge if their similarity value is higher than a predefined threshold and projected using the classical Fruchterman-Reingold algorithm. The main feature of the tool is the calculation and visualization (through color) of a "cliff index", which describes whether the compound has a big shift in potency compared to its neighborhood. HiTSEE can also help in the detection of activity cliffs by spotting big changes in size within a given cluster. While SARANEA is more targeted towards the exploration of a full set of compounds, HiTSEE leverages on the idea of having a small set of initial compounds (often a single one) and to explore their direct neighborhood.

Scaffold Hunter [5] is another visualization tool that can be used to find relations between structural features and activity level. A tree structure is built starting from one compound and building molecular scaffolds by removing rings in the periphery, through a series of chemistry and medicinal chemistry rules. The visual representation is a radial tree depicting the hierarchical scaffold structure and the activity level of the scaffolds. The tool can be used to investigate the potential of the scaffold to be at the origin of activation of biological processes. Together with HiTSEE, Scaffold Hunter shares the idea of starting the analysis from one (or more in HiTSEE) compounds of interest and exploring their neighborhood. However, HiTSEE is focused more on structural similarities between the compounds rather than the scaffolds.

DrugViz [6] is a newly developed plug-in in the Cytoscape environment in which analysis is centered on a network representation of interactions between biological targets. The system enables similar targets to be picked and common compounds to be found, or alternatively, the system selects similar compounds and sees how they are distributed in the target network. While the system allows the investigation of relationships between biological targets and molecular structures, visualization is not targeted towards the visualization of structure-activity relationships.

The SAR Map [7] (and its extension Enhanced SAR Map) allows the focus to lie on one single molecule of interest, exploring all its substituents through R-group analysis. For its input, R-group analysis takes a list of compounds with a common scaffold and generates all possible variations. The SAR map is a heat map where the columns and the rows represent substituents of two selected variation sites. Each cell represents one specific compound (formed by attaching the substituents) and a color map, or more complex visualizations provide rich information about each compound. HiTSEE also allows the variations around a subset of structurally similar compounds (in the molecule details view) to be visualized, however this functionality takes place in the larger context of similarity analysis among entire molecules.

ChemGPS-NP [8], similarly to HiTSEE, projects molecules in a low-dimensional space using a PCA projection. The visualization is designed in a way so as to reflect those properties that are relevant for bioactivity. However, the visualization does not directly adress the correlation between activity and structural similarity.

In summary, HiTSEE has the unique advantage of allowing flexible and smooth navigation in the chemical space by conjugating two contrasting needs: the need to create visual summaries of chemical libraries and the need to explore the neighborhood of selected compounds.

Data processing & quality control

The researcher normalizes the data against the control values and analyzes the result to check for abnormal behaviors. A number of biases and outliers can exist in an HTS experiment. In our environment we typically check for assay plates with a low Z-factor [9], a measure of assay quality based on the controls. This phase is also supported by a visual tool we have developed previously [10] to visually explore the values in the plates. The researchers can directly filter out those compounds with unreliable readouts or simply mark them for future analysis.

Hit selection

The goal of this phase is to identify the compounds that reacted in the experiment. Typically, the researcher organizes the results in a list sorted according to activity level and chooses a threshold value above which the compounds are considered active. These compounds are called hits.

Hit confirmation

After hit identification, the researcher re-tests the selected hits in a new and more focused screening (e.g. testing different concentrations of the compounds and calculating the half maximal inhibitory concentration (IC₅₀)) in which confirmation of activity is sought. This phase is normally done manually given the small number of molecules to test.

Hit exploration

After hit confirmation the researcher explores the chemical space around the hits. Typically, he or she is in search of relationships between molecular structure and activity to isolate molecular fragments that induce activity (a process called structure-activity relationship (SAR) analysis).

Hit expansion

Similar to hit exploration, hit expansion focuses on exploring the space around the hits but the focus is on identifying alternative molecules that retain the desired properties and meet additional requirements (e.g. solubility).

Each one of these steps can involve data and computational resources. While in our environment we provide support for all these stages, HiTSEE provides only support for a subset of them, namely: selection, exploration, and expansion.

Data processing and tasks

In the following we provide additional details about the data and describe how it is processed before entering into the description of HiTSEE. Then, we describe how we gathered the requirements for the design of the tool and discuss our main motivation for focusing on a subset of the HTS tasks.

HiTSEE

HiTSEE's interface is organized around three main views: list+projection view (Figure 1 (left, middle)), molecules detail (Figure 1 (right)) and substructure search view (Figure 2), that support exploration, in-depth investigation and structural queries. The list+projection view permits molecules of interest to be selected and to project them in a scatter plot visualization to form clusters of (structurally) similar compounds. The view supports the investigation of relationships between activity levels, structural features, and other chemical properties. The molecules detail and substructure search view shows the molecular structure of compounds selected in the projection view and triggers substructure searches.

In the following we describe each component in detail together with the interaction capabilities offered by each one.

List+Projection view

The List+Projection view consists of two interactive elements: a compounds list and a linked scatter plot view. (Figure 3 (left)) The compound list organizes the full set of compounds in the library in a list format sorted by activity level. Each item is represented by its molecular structure and by a bar with length proportional to its activity level.

The user can select one or more items in the list, project them in the scatter plot view, and expand the selection to a user-defined number of neighbors. The neighbors are the compounds that are structurally most similar to the current selection. The structural similarity is calculated from the fingerprint bit-vectors (see Fingerprint Section).

The compounds are represented by circles and positioned in the view through a multidimensional scaling (MDS) projection such that compounds with similar structures occupy similar positions. Size represents the activity level and color is used to distinguish between those compounds included in the initial selection and those added by the expansion mechanism. Each circle also contains a small modified pie chart representing additional chemical properties of interest (in our case the LogP value). The pie chart is designed in a way to turn its fill color into a more prominent one (darker blue) when the value of interest goes beyond a predefined threshold.

The MDS projection takes a distance matrix of metric distance values as input. For each pair of compounds, we calculate the Tanimoto [17] distance between their fingerprint bit-vectors. Two problems emerge from MDS-based projections: overlapping items and fuzzy boundaries between the groupings. To cope with these two issues we implemented two additional features. First, we used an overlap removal mechanism that permits to displace the items from their original position if they overlap each other. Second, in order to facilitate the grouping of the items, we cluster the items and draw a "bubble" around them to reinforce the perception of grouping. The clustering algorithm takes the screen-space positions of the items as input and clusters them into bubble sets [18]. For each cluster, we determine the common substructure of all containing compounds and position it left to the cluster.

In designing the projection view, we tried to optimize its visual effectiveness towards reading patterns with biological interest. In the following we provide a summary of the rationale behind our main design choices.

Since position is the visual variable that can be perceived pre-attentively most accurate [19], we use it to convey molecular similarity (through the proximity data given by MDS), which is the most important piece of information in the data. Activity level is mapped to circle size (with a square root mapping to take into account the area effect) to allow for easy discrimination among the molecules. While visual variables like bar length allow for a more accurate comparison of values [19], we decided to use circles and their size because: (1) they cluster more naturally than shapes with other aspect ratios, (2) they are more robust w.r.t. the overlapping removal mechanism, (3) they allow for easy discrimination between high vs. low activity molecules while keeping the visualization compact, (4) reading the activity values accurately is not the main purpose of the visualization (as long as major differences can be spotted). A third parameter (LogP) is encoded as the visual variable angle. To allow better readability we visualize the angle by using a filled pie chart with only one pie embedded in the circles. While a number of alternatives exist to encode two parameters, as for instance stacked bars and nested circles (see Figure 3), we decided to use a modified pie chart because it corresponds well with the circular shape we adopted and readability scales visually better than nested rings to items of different size.

HiTSEE for KNIME

Integration of HiTSEE into the KNIME platform is achieved by the development of a series of processing nodes that realize the functions developed in HiTSEE and additional helper nodes that permit to build a fully functional pipeline. Figure 4 shows an example of the workflow. The data can be read and pre-processed using the large set of available nodes in the KNIME platform (many specifically built for processing biological data). In the example, the first set of nodes permits to load the data into the system; preprocess it to calculate derived information, such as logP and fingerprints; and to apply some normalization functions. The nodes between the "Loop Start" and "Loop End" represent the core functionalities offered by the HiTSEE KNIME integration and reflect the main behavior of HiTSEE as a continuous iteration loop to focus on a specific set of compounds.

Each iteration starts with the selection of a molecule subset from the user directly or indirect from a user defined search request. The first set of nodes (b) is responsible for the generation of the expanded set (for a project-and-expand operation) and for the search of the target molecules of a substructure search. The second set of nodes (c) is responsible for projection and clustering. More specifically, this node implements the following set of functions: (1) calculate the distances between the molecules in the selected (and expanded) subset; (2) project the distances; (3) cluster the projected positions; (4) find a most common substructure as cluster label. Finally, the last node (d) is the one that implements the visualization, the interactive functions, and the loop control.

The "Project" and "Project and Expand" buttons trigger the HiTSEE loop and permit it to execute one iteration of the steps described above under the parameters given by the user. The main view is split into the previously described list view, projection view, and detail view. The example in Figure 1 shows a selection made by the user (highlighted in gray). When the "Project & Expand" button is clicked, the following parameters are set for the next loop cycle. The three selected molecules are the new subset of interest, an expansion of the subset is required for a set size of 15 additional compounds. The loop is started with these parameters for one cycle and the main view changes accordingly.

Case study 1: the Kif18A data set

We used HTS-data generated by a screen looking for a specific inhibitor of Kif18A [21] to prove the effectiveness and usability of HiTSEE.

Kif18A belongs to the family of mitotic kinesins. Kinesins are ATP dependent motor proteins, which utilize the energy derived from ATP hydrolysis to produce mechanical force. Kif18A belongs to the kinesin-8 family whose members are known to be required for the correct segregation of chromosomes in mitosis. Besides its key function in mitosis, Kif18A is characterized by its unique enzymatic properties since it integrates both motility and microtubule depolymerization activity. Due to its central function in mitosis and intriguing enzymatic properties we applied a small molecule screen to identify small molecules that inhibit the ATPase activity of Kif18A. The published results of the small molecule screen [21] can be applied as a proof-of-concept principle to validate HiTSEE.

Case study 1: hit selection

The first step after a high throughput screen is to decide which positive results are counted as a hit, whereupon the compounds are sorted according to their activity level. With HiTSEE the compounds are directly sorted according to their activity level and the corresponding structures are represented in a list. After choosing the 30 compounds with the highest activity and projecting them, we have a first set for hit selection. In the projection we could see the common structure of the clusters (Figure 5a). The only common motif in this case was the phenyl moiety, which is not really significant. Also blue filling of the dots indicates a LogP value above 5, which could cause solubility problems in aqueous media. Nevertheless, by removing the overlap we are able to see the structure of the active compounds and get first hints for structures relevant for activity. In the detailed view of all these selected compounds we can easily compare their structures by eye and spot new common or interesting structures like the diphenyl sulfide moiety. With the project and expand option we were able to see structures related to the hits with lower activity levels (indicated by orange dots, Figure 5b). Clusters of highly active and less active compounds made us feel more confident to select the highly active compound as a hit, because the structural motif is spread over an activity range.

Case study 1: hit expansion

After one hit has been selected we can choose structural elements and search the entire library for that motif. In our case we started by projecting the 30 compounds with the highest activity in the view. After selecting all the hits and going through the detailed view we decided to look for compounds containing a diphenyl sulfide moiety, which was present in several hits (Figure 6). The result was that we received nice clusters of compounds. This clustering gives us more confidence to choose the selected hit for further testing because now we know that there are other compounds containing the same structural moiety with a different activity level. This activity range allows us to establish SARs, identify the core structure necessary for activity and hint to structures with higher activity. Further this finding appears to indicate that the selected hit is not a false positive because the substructure is present in different compounds. Using the detailed view of a cluster, a list of compounds with common substructures and diverse activity levels is obtained. The fact that the common motif is not only present in the highly active compounds but also in less active ones enabled us to establish SARs first and to feel confident in choosing this compound for further investigations. The search did not give BTB-1 as a result but it gave a whole set of BTB-1 analogues, which are also very active. If we were to continue with these results we would choose the diphenyl sulfone moiety as the lead structure, the published Kif18A inhibitor [21]. After investigating compounds with this structural motif and establishing SARs we would finally end at BTB-1. HiTSEE allows hits to be confirmed by exploring the chemical space around them and revealing less active compounds bearing same structural moieties.

Case study 2: NCI AIDS antiviral screen

The NCI AIDS Antiviral Screen data set [22] provides measures of the ability of chemical compounds to protect cells against the cytopathic effects of HIV (see Weislow et.al. [23]). The library consists of more than 40000 compounds as well as chemical structure descriptions and screening results.

The 20 compounds that showed the highest activity level and high diversity were automatically preselected and projected in order to obtain an initial set of structures for hit selection. Within the projection view we get different clusters. By hovering over items we could see the chemical structure of each single compound. A cluster containing different heterocyclic molecules was selected and the detail view was used in order to find common substructures. The detailed view revealed a set of compounds containing benzothiazoles, quinazolines and quinolines, which are related structures. In order to confirm the selected compounds as active hits the project and expand option was used and the found cluster was investigated for common structures in the detailed view of HiTSEE. Several compounds bearing a quinoline scaffold were identified with different activity levels (Figure 7). Therefore we would suggest using this scaffold as a true hit and as a lead structure for further synthesis to establish SARs. Even though the confirmed active compounds of the published results of the screen did not list quinolines as active compounds, the related N-heterocyclic benzodiazepines, thiazolobenzimidazoles and pyrimidne analogues were confirmed active hits. Further details and the mode of action of these compounds can be found in De Clerq [24].