Visualizing chemical structure-subcellular localization relationships using fluorescent small molecules as probes of cellular transport
© Rosania et al.; licensee Chemistry Central Ltd. 2013
Received: 18 June 2013
Accepted: 1 October 2013
Published: 5 October 2013
To study the chemical determinants of small molecule transport inside cells, it is crucial to visualize relationships between the chemical structure of small molecules and their associated subcellular distribution patterns. For this purpose, we experimented with cells incubated with a synthetic combinatorial library of fluorescent, membrane-permeant small molecule chemical agents. With an automated high content screening instrument, the intracellular distribution patterns of these chemical agents were microscopically captured in image data sets, and analyzed off-line with machine vision and cheminformatics algorithms. Nevertheless, it remained challenging to interpret correlations linking the structure and properties of chemical agents to their subcellular localization patterns in large numbers of cells, captured across large number of images.
To address this challenge, we constructed a Multidimensional Online Virtual Image Display (MOVID) visualization platform using off-the-shelf hardware and software components. For analysis, the image data set acquired from cells incubated with a combinatorial library of fluorescent molecular probes was sorted based on quantitative relationships between the chemical structures, physicochemical properties or predicted subcellular distribution patterns. MOVID enabled visual inspection of the sorted, multidimensional image arrays: Using a multipanel desktop liquid crystal display (LCD) and an avatar as a graphical user interface, the resolution of the images was automatically adjusted to the avatar’s distance, allowing the viewer to rapidly navigate through high resolution image arrays, zooming in and out of the images to inspect and annotate individual cells exhibiting interesting staining patterns. In this manner, MOVID facilitated visualization and interpretation of quantitative structure-localization relationship studies. MOVID also facilitated direct, intuitive exploration of the relationship between the chemical structures of the probes and their microscopic, subcellular staining patterns.
MOVID can provide a practical, graphical user interface and computer-assisted image data visualization platform to facilitate bioimage data mining and cheminformatics analysis of high content, phenotypic screening experiments.
KeywordsMachine vision Cheminformatics Virtual reality Data mining Optical probes Multivariate analysis Human-computer interaction Graphical user interface
Combinatorial libraries of prospective, organelle-targeted small molecule fluorescent probes have served as treasure troves of optical sensors of cell physiology [1–11]. Until recently, development of live cell imaging probes relied mostly on qualitative, manual microscopic observations of stained cells. With the invention of high content screening systems [12–18], microscopic imaging experiments can be automated, so that image data sets can be acquired from cells incubated with hundreds or thousands of candidate bioimaging probes, with the click of a button [5, 14, 16, 17, 19, 20]. Nevertheless, the ability to identify trends in staining patterns across large numbers of probes and to link these trends to specific variations in the structure or physicochemical properties of the chemical agents has remained a major challenge. For bioimaging probe discovery and optimization efforts, data acquisition has been greatly facilitated by advances in combinatorial chemical synthesis and robotic screening technology as well as machine vision and cheminformatics analysis. Nevertheless, for bioimaging probe development, the ability to interpret machine vision and cheminformatics results based on visually recognizable, cellular staining patterns assessed by human viewers is essential.
Integrating advances in the cognate fields of cheminformatics and bioimage informatics [15, 21–23], our research group has been developing computational tools to address the data mining challenges inherent to complex phenotypic screening experiments, such as bioimaging probe discovery and development efforts [6, 24–27]. In 2003, we began cell-based screening experiments involving a combinatorial libraries of cell permeant, organelle-targeted fluorescent compounds . For data mining, we began integrating machine vision [13, 27–30] and cheminformatics approaches [13, 24, 27–31]. We demonstrated the ability to link the different chemical building blocks of small molecule bioimaging probes to the resulting staining patterns by applying multivariate regression approaches [24, 28, 29]. Independently, we developed predictive models linking the physicochemical properties of small molecule chemical agents to their subcellular distribution [30, 32–34] to facilitate discovery and development of in vivo, site-directed bioimaging probes [35, 36].
In spite of all these advances, the human observer still plays an important role in the discovery of new classes of bioimaging probes [27, 29] because ultimately, staining patterns have to be recognizable, interpretable and mechanistically meaningful to the human eye. Furthermore, without visual validation, unsupervised image analysis can be affected by many potential imaging artifacts including focus drift, motion blur or the presence of insoluble precipitates or aggregates, which affects the interpretation of cell-associated staining patterns. On their own, machine vision algorithms may lead to misleading or erroneous interpretations because of inaccurate object segmentation, image registration, and other sources of error. Therefore, to facilitate visual inspection of large image data sets of probe screening experiments, we decided to construct MOVID: a practical, cheminformatics and machine vision-assisted, virtual reality-based image visualization platform, built from off-the-shelf hardware and software components. Here, we demonstrate how MOVID can be used to interpret the staining patterns of a combinatorial library of small molecule fluorescent probes of cellular transport, relating these patterns to the chemical structure, physicochemical properties and the predicted subcellular localization patterns of the probes.
Acquisition of image data sets
Image data was obtained from HeLa cells incubated in the presence of 1,344 fluorescent styryl compounds synthesized by conjugating one of eight pyridinium or quinolinium groups (A-H) to one of 168 aldehyde groups (1-168) [27, 28]. All 8 × 168 combinations were screened using a Kineticscan™ instrument (Cellomics, Inc., Pittsburgh, PA) to acquire images using the 20X objective of the instrument [27, 28]. The cell permeant DNA stain Hoechst™ 33342 (“Hoechst”) was included in the media, as an orthogonal reference marker which labeled the cell nuclei to allow segmenting the individual cells in the images [27, 28]. From each probe, twelve images were obtained. Six images were acquired after one hour incubation with the probe, and another six images where acquired after the probe was removed from the extracellular medium. The images were 512 × 512 pixels, with pixel intensities ranging from 0 to 4,095, corresponding to Hoechst, TRITC, FITC, and Cy5 channels (TRITC and FITC channels images were acquired with two camera exposure settings). For visualization, only the Hoechst™ and the 1 second exposure FITC, TRITC and Cy5 images were displayed. This image dataset was deposited in an open, public repository (http://deepblue.lib.umich.edu/; Query “HeLa cells incubated with styryl compound”).
Calculation of quantitative image features associated with cellular staining patterns
Machine vision was used to analyze the cellular staining patterns of the styryl probes [27, 28]. Briefly, the Hoechst channel images were used to identify the nuclear region of each cell. The resulting nuclear regions were then dilated by 10 pixels to create a cell mask. For analyzing styryl compound fluorescence, the images acquired with the FITC, TRITC and Cy5 channels were summed to create a total cellular intensity multichannel image corresponding to the signal of the styryl molecule fluorescence over the multiple acquisition channels . Using the complement of the dilated cell region mask, the background pixel intensity of the multichannel image was measured. The median intensity value of the background pixels of each image was subtracted from the intensity value of every pixel in that image, and these adjusted intensity values were then truncated at zero. Images lacking styryl-specific fluorescence signal (e.g. with cellular fluorescence signal at or below the extracellular background fluorescence, or showing extensive saturation of pixel intensities) were excluded from further analysis. With images exhibiting above background cell-associated fluorescence, three key cell-associated image features capturing the optical signal intensity and staining patterns of the probes were measured [27, 29]: 1) the ratio of cell associated pixel intensity/background pixel intensity; 2) the cytoplasmic-to-nucleus ratio of cell-associated pixel intensities; and, 3) the coefficient of variation of cell-associated pixel intensities.
Cheminformatics analysis of structural, topological, and physicochemical property features
Physicochemical properties of styryl compounds and their building blocks were calculated with MOE (Molecular Operating Environment, Chemical Computing Group Inc., Montreal, Canada). Diversity with respect to quantitative properties was assessed by comparing histograms for each property between the styryl compounds and a reference dataset of hundreds of chemical agents with published subcellular localization features, compiled from a literature-based review surveying the organelle-targeting properties of small, drug-like molecules [30–32]. Diversity of chemical structures for the styryl molecules and for their chemical building blocks was assessed using Cactvs bit strings which represent chemical structures as 881- dimensional binary sequences. Each bit in a Cactvs sequence indicates the presence of a specific chemical substructure in the overall structure. The Tanimoto coefficient (Tc) was used to quantify the similarity between each pair of structures, by dividing the number of features in the Cactvs string that are common to both structures by the total number of features present in either structure. Mathematically, Tc = C/(N1 + N2 - C) where N1 and N2 correspond to the number of features present in the fingerprints of each molecule in the pair, and C is the number of features present in the fingerprints of both molecules.
Predicting subcellular distribution patterns
The theoretical distribution of styryl molecules inside cells was calculated using a published mathematical model [30, 33, 37], based on the predicted passive diffusion properties of the compounds in the presence of a constant extracellular concentration of the compound. This mathematical model incorporates Fick’s law of diffusion, Henderson-Hasselbalch equation and the Nernst-Planck equation, to model the diffusion of small molecule chemical agents across the membranes bounding cellular / subcellular compartments, as determined by the concentration gradients, pH gradients, and electrical potentials across the bounding membranes. Computationally, the models allowed calculating the mass of compound that accumulates in various subcellular compartments, over time [30, 33, 34, 38]. For simulating the transport behavior of the styryl molecules, all styryl molecules possess a single, fixed positive charge plus 0 or more additional ionizable groups. For this study, simulations were done only for molecules with 0 or 1 additional ionizable groups corresponding to 1256 out of 1344 molecules in the library (93.4%). Baseline input parameters of the model were based on an “average” mammalian cell using the established, generic physiological parameters of cells (such as membrane surface area, cellular and organelle volumes, pH values in each compartment, membrane potentials, and lipid fractions in each compartment) [33, 39]. Input physicochemical properties of styryl molecules (such as logP and pKa) were calculated with Chemaxon (http://www.chemaxon.com). After the concentration at steady state was calculated for each compartment, the mass of styryl molecules in each compartment was calculated by multiplying the concentration in that compartment by the compartment’s volume. The total mass of compound in cell was then calculated as the sum of the masses in the different compartments: cytoplasmic (or plasma) membranes, lysosomes and mitochondria. The fractional mass of compound in cytoplasmic (or plasma) membranes, lysosomes and mitochondria was calculated based on the mass of compound in each compartment divided by the total mass of compound in these three compartments, at steady state.
Developing machine vision- and cheminformatics-assisted MOVID platform
Analyzing the chemical diversity of the organelle-targeting styryl library
Characterizing the diversity of staining patterns in relation to the diversity of predicted localization patterns
Visualizing relationship between chemical similarities and staining pattern similarities
To demonstrate this analysis, we focused our attention on a styryl probe that was previously found to accumulate in mitochondria [44–46] and is one of the best characterized probes in the styryl library (Figures 6 and 7, upper left). To demonstrate how chemical structure variation relates to variation in staining patterns, we identified two subsets of compounds in the styryl library – one consisting of compounds that have similar structures relative to reference probe (Figure 6), and one consisting of compounds that have dissimilar structures relative to reference probe (Figure 7). The subset of compounds that are structurally similar to the reference probe consisted of those compounds in the library having both the pyridinium or quinolinium building block and the aldehyde building block with a Tanimoto coefficient >0.4 relative to the reference probe. By visual inspection, the staining patterns of these probes appeared similar to the organelle-associated staining pattern of the reference probe, although there was substantial variation in the overall intensity (Figure 6). The subset of compounds deemed least similar to the reference probe consisted of compounds whose aldehyde building block had a Tanimoto coefficient <0.25 relative to the aldehyde building block of the reference probe. For these compounds, the staining patterns appeared to be different from the staining patterns of the reference probe (Figure 7). By visual inspection, the staining patterns of the probes revealed roughly similar numbers of membrane- and organelle- staining probes, and a smaller fraction of nuclear-staining probes. Based on the Tanimoto coefficients, the transition between visually similar vs. different staining patterns in relation to the Tanimoto coefficients appeared to be abrupt and nonlinear, with most compounds possessing an aldehyde group with a Tanimoto coefficient >0.4 exhibiting similar staining patterns as the reference probe (Figure 6). This observation may be an explanation for why previous, unsupervised machine vision and cheminformatics analyses did not reveal any significant correlation between Tanimoto coefficients and quantitative image features associated with the staining patterns .
Visualizing relationship between staining patterns and regressed contributions of building blocks to staining patterns
In this study, we integrated the results of machine vision and cheminformatics analysis, with an image data visualization approach (MOVID), to facilitate visual interpretation of quantitative relationships between the chemical structure of a combinatorial library of small molecule fluorescent probes of cellular transport and their associated, cellular staining patterns captured in a large image data set. With MOVID, the ability to visualize multidimensional relationships was enhanced by allowing human viewers to literally navigate through sorted, high resolution image data sets, with the aid of an avatar. While at the level of the entire image, the ability to resolve local subcellular staining patterns is lost (Figure 4B, E, H), zooming into the individual images revealed details of the fluorescence signal in nucleus and cytoplasm of the individual cells (Figure 4A, D, G) that could be interpreted in the context of the chemical structure, physicochemical properties, and predicted localizations of the molecules. In this manner, MOVID facilitated interpretation of complex relationships between the phenotypic staining patterns of the cells and the chemical properties of the probes.
Using MOVID, we related the predicted, percent mass distribution of styryl molecules in different subcellular compartments to their observed staining patterns. When viewed across a large number of probes ranked based on their percent mass distribution in lysosomes, mitochondria and cytoplasmic (or plasma) membranes, a strict correspondence between predicted and visual localization patterns was not readily apparent (Figure 3). To further explore this relationship between the predicted localization patterns and observed staining patterns, the images obtained with the styryl compounds were visually classified by an expert observer (Figure 4). These observed staining patterns were used to calculate the accuracy of the predictions (Figure 5). Results of this analysis confirmed the original, qualitative visual impression (Figure 3), in that the predicted mass distribution of the probe in different compartments generally did not correspond with specific staining patterns. Only those compounds predicted to have a mitochondrial or lysosomal mass distribution of 40% or less showed a consistent association with the expected, nonmitochondrial and nonlysosomal staining pattern.
In addition, MOVID allowed us to confirm previous results obtained through machine vision and cheminformatics analyses [27, 28]. For example, by visual inspection, similarities in the calculated physicochemical properties and subcellular distributions of the probes did not correspond to similarities in their subcellular localization patterns captured in the images (Figures 6, 7). Only compounds with Tc < 0.25 in the aldehyde building block (Figure 6) yielded clear visual differences in staining patterns as compared to the reference probe, explaining the absence of quantitative correlation between variations in Tc and variations in image features . Furthermore, in the regression analysis, visual inspection of cellular staining patterns (Figure 8) also confirmed previous machine vision results  in terms of the ability of aldehyde and pyridinium or quinolinium building blocks to exert a predictive, additive contribution to diffuse vs. organelle-associated staining patterns, apparent across large numbers of images.
In light of previous machine vision and cheminformatics analyses [24, 28], we also noted that the range in chemical diversity encoded by the variations in pyridinium or quinolinium building blocks was less than the range of chemical diversity encoded by variations in aldehyde building blocks (Figure 2). This corresponded to a more limited range in of regressed contributions to the staining patterns. Visual inspection of image arrays with MOVID were consistent with previous quantitative image analysis studies , indicating that chemical variations in aldehyde building blocks exerted a similar effect on the staining patterns as compared to chemical variations in the pyridinium or quinolinium building blocks. Previously, quantitative image analysis indicated that a charge push-pull mechanism affecting the electron distribution between the imminium group of the pyridinium or quinolinium and an amine group present in the opposite side of the molecule could exert a strong effect on quantitative image features . Observed trends spanning large numbers of images were consistent with electron migration across the central methine bridge of the styryl molecule being associated with the diversity of staining patterns exhibited by styryl molecules. Interestingly, logP calculations generally do not take the effects of charge migration into account, which could partly explain the discrepancies between the predicted distribution of the probes and the observed staining patterns, with the exception of nonmitochondrial and nonlysosomal localizing compounds (Figures 3, 4 and 5).
In terms of the integration of off-the-shelf hardware and software components, the virtual, graphical user interface of MOVID offered distinct advantages over larger multipanel LCD arrays for high resolution image data display, also known as “command and control centers” or “data walls” (e.g. http://www.bioimage.ucsb.edu/iWall). MOVID was a practical, user-friendly and ergonomic alternative to display large amounts of visual information at high resolution, over a large field of view, to an individual user sitting at a desk. With the aid of an avatar, a viewer was able to rapidly scan through large numbers of images at very high resolution from a very close (virtual) distance, or at lower resolution from a farther (virtual) distance away (Figure 1). While sitting at a desktop computer workstation, a user can switch between MOVID to cheminformatics or modeling software with the click of the mouse (Figure 1A). In the past, many health care simulation, crowd-sourcing, open science, scientific conferencing and educational applications have been facilitated by the multiuser capabilities of online, open source virtual reality browsers [40–43]. Similarly, the open virtual reality graphics and multiuser interface capabilities of MOVID could also facilitate simultaneous, multiuser visualization capabilities at minimum cost.
To conclude, the integration of cheminformatics and machine vision with a high resolution, multidimensional virtual reality image display and graphical user interface helped us overcome a major bottleneck we encountered while attempting to analyze the results of a phenotypic cell based assay. The associated data visualization and interpretation challenges are exemplified by the analysis of chemical determinants of the subcellular distribution properties of fluorescent small molecules probes of cellular transport, as elaborated in this study. In particular, despite their limited chemical diversity, our data was consistent with styryl molecules displaying a broad range of localization features, including probes that selectively labeled nuclei , nucleoli , plasma membrane  as well as other components [48, 49]. Because bioimaging data is ultimately interpreted by human observers, visualization remains an essential component of bioimaging probe discovery and development, Nevertheless, beyond bioimaging probes, MOVIDs should be broadly applicable to many other phenotypic screening experiments [5, 8, 10, 20, 50, 51] and subcellular drug targeting and delivery studies [2, 18, 33, 39].
Visual scoring of subcellular localization patterns
To study how the predicted subcellular distribution patterns of the styryl molecules were related to their observed staining patterns, a total of 199 distinct images possessing a visible fluorescence signal were visually scored by an expert human observer, based on three distinct categories: (1) mitochondrial or lysosomal staining pattern; (2) cytoplasmic (or plasma) membranes staining pattern; (3) nuclear or nucleolar staining pattern. Only images of compounds that could be unambiguously classified in terms of one of these three distinct categories were considered. In some images, rounded (detached, mitotic or dying) were also present, and the localization of the probe in those cells was ignored. Images containing only rounded cells were not included in this analysis. This resulted in a subset of 123 images that could be used to compare how the predicted localizations related to the visual staining patterns.
Relating the predicted, mitochondria- or lysosome-targeting styryl molecules to their visual staining patterns
For the mitochondrial or lysosomal localization predictions, the Accuracy of the predictions was plotted as a function of continuous thresholds of predicted percent mass distribution of the compounds in mitochondria or lysosomes, in the range of 0 to 100%. Independently, for the cytoplasmic (or plasma) membrane predictions, the Accuracy of the predictions was plotted as a function of continuous thresholds of, predicted percent mass distribution of compounds in mitochondria or lysosomes in the range of 0 to 100%.
Relating the predicted, cytoplasmic (or plasma) membrane-targeting styryl molecules to their visually-scored staining patterns
Independently, we also calculated the accuracy of cytoplasmic (or plasma) membrane localization predictions in relation to the observed, visual staining patterns. For images of probes that were visually scored as possessing a cytoplasmic (or plasma) membrane localization pattern, we first established whether the calculated, percent mass distribution of the probe in cytoplasmic (or plasma) membranes was (i) equal to or greater than a threshold calculated percent mass distribution of probe in the cytoplasmic (or plasma membrane) compartment (true positives); or ii) lower than a threshold calculated percent mass distribution of probe in cytoplasmic (or plasma) membrane compartment (false positives). In parallel, we established whether the calculated, percent mass distribution in cytoplasmic (or plasma) membranes for each probe observed as nuclear, nucleolar, mitochondrial or lysosomal staining was (iii) equal to or greater than the threshold calculated percent mass distribution in cytoplasmic (or plasma) membrane (false negatives); or (iv) less than the threshold calculated percent mass distribution in cytoplasmic (or plasma) membranes (true negatives). In turn, the Accuracy of the cytoplasmic (or plasma) membrane distribution predictions was assessed in relation to their observed visual staining patterns based on the number of correctly classified and incorrectly classified compounds, in an analogous manner as was done for the mitochondrial or lysosomal distribution predictions (using Eq. 1), and plotted as a function of continuous thresholds of predicted, percent mass distribution in mitochondria or lysosomes in the range of 0 to 100%.
Visualizing relationships between building blocks and staining patterns
To visualize relationship between similarities in the chemical structure of the building blocks and the associated images acquired from styryl molecules synthesized from those building blocks, image data sets were sorted in 2-dimensional arrays, based on the ranked, pair-wise Tanimoto coefficients between aldehyde and pyridinium or quinolinium building blocks, relative to the building blocks of a single, reference compound. The building blocks were sorted, in ranked order of decreasing pair-wise Tanimoto coefficients, relative to the building blocks of the reference probe. With MOVID, the image associated with the reference probe was displayed in the upper left corner of the image arrays. The individual images acquired from the styryl probes (made from pair-wise combinations of aldehyde and pyridinium or quinolinium building blocks) were displayed in the row and column position corresponding to the rank-ordered Tanimoto coefficient of the respective building blocks, relative to the reference probe.
Visualizing the contribution of the different building blocks to the resulting staining patterns
The extent to which the two building blocks of the styryl molecules were associated with the various measured image features was calculated using a multivariate regression approach . Briefly, in the styryl library, each molecule is comprised of an aldehyde building block that is chemically conjugated with a pyridinium or quinolinium building block. Thus, quantitative image features characterizing the cell-associated intensity and spatial signal of bioimaging probes can be related to each building block, assuming a simple additive contribution of each building block to any measurable image feature. For example, for any given image feature Y (Y = total cellular intensity, or N/C ratio, or nuclear CV), the linear model f(Y) = A(i) + P(j) can fit, where f is a suitable transformation such as the logarithm or identity function, i is the index of the aldehyde building blocks (A), and j is the index of the pyridinium or quinolinium building blocks (P). For visualization, aldehyde and pyridinium/quinolinium building blocks were sorted based on their contribution to the fitted numerical image features. Images were displayed with MOVID, with the sorted aldehyde building blocks corresponding to the rows in the grid, and the sorted pyridinium/quinolinium building blocks corresponding to the columns.
Multidimensional online virtual image display
Liquid crystal display
Texas Red Isothiocyanate.
We thank Lilly Roy and Maria Posada for their help setting up the MOVID platform. This work was supported by NIH grant R01GM078200 to GRR.
- Ljosa V, Carpenter AE: High-throughput screens for fluorescent dye discovery. Trends Biotechnol. 2008, 26 (10): 527-530. 10.1016/j.tibtech.2008.06.008.View ArticleGoogle Scholar
- Kelly KA, Shaw SY, Nahrendorf M, Kristoff K, Aikawa E, Schreiber SL, Clemons PA, Weissleder R: Unbiased discovery of in vivo imaging probes through in vitro profiling of nanoparticle libraries. Integr Biol (Camb). 2009, 1 (4): 311-317. 10.1039/b821775k.View ArticleGoogle Scholar
- Giuliano KA, Taylor DL, Waggoner AS: Reagents to measure and manipulate cell functions. Methods Mol Biol. 2007, 356: 141-163.Google Scholar
- Zhai D, Lee SC, Vendrell M, Leong LP, Chang YT: Synthesis of a novel BODIPY library and its application in the discovery of a fructose sensor. ACS Comb Sci. 2012, 14 (2): 81-84. 10.1021/co200136b.View ArticleGoogle Scholar
- Wagner BK, Carrinski HA, Ahn YH, Kim YK, Gilbert TJ, Fomina DA, Schreiber SL, Chang YT, Clemons PA: Small-molecule fluorophores to detect cell-state switching in the context of high-throughput screening. J Am Chem Soc. 2008, 130 (13): 4208-4209. 10.1021/ja077656d.View ArticleGoogle Scholar
- Li Q, Kim Y, Namm J, Kulkarni A, Rosania GR, Ahn YH, Chang YT: RNA-selective, live cell imaging probes for studying nuclear structure and function. Chem Biol. 2006, 13 (6): 615-623. 10.1016/j.chembiol.2006.04.007.View ArticleGoogle Scholar
- Lee JS, Vendrell M, Chang YT: Diversity-oriented optical imaging probe development. Curr Opin Chem Biol. 2011, 15 (6): 760-767. 10.1016/j.cbpa.2011.10.007.View ArticleGoogle Scholar
- Vendrell M, Lee JS, Chang YT: Diversity-oriented fluorescence library approaches for probe discovery and development. Curr Opin Chem Biol. 2010, 14 (3): 383-389. 10.1016/j.cbpa.2010.02.020.View ArticleGoogle Scholar
- Lee JS, Kim YK, Vendrell M, Chang YT: Diversity-oriented fluorescence library approach for the discovery of sensors and probes. Mol Biosyst. 2009, 5 (5): 411-421. 10.1039/b821766c.View ArticleGoogle Scholar
- Finney NS: Combinatorial discovery of fluorophores and fluorescent probes. Curr Opin Chem Biol. 2006, 10 (3): 238-245. 10.1016/j.cbpa.2006.04.025.View ArticleGoogle Scholar
- Mello JV, Finney NS: Reversing the discovery paradigm: a new approach to the combinatorial discovery of fluorescent chemosensors. J Am Chem Soc. 2005, 127 (29): 10124-10125. 10.1021/ja043682p.View ArticleGoogle Scholar
- Giuliano KA, Johnston PA, Gough A, Taylor DL: Systems cell biology based on high-content screening. Methods Enzymol. 2006, 414: 601-619.View ArticleGoogle Scholar
- Chen VY, Khersonsky SM, Shedden K, Chang YT, Rosania GR: System dynamics of subcellular transport. Mol Pharm. 2004, 1 (6): 414-425. 10.1021/mp049916t.View ArticleGoogle Scholar
- Giuliano KA, Haskins JR, Taylor DL: Advances in high content screening for drug discovery. Assay Drug Dev Technol. 2003, 1 (4): 565-577. 10.1089/154065803322302826.View ArticleGoogle Scholar
- Shariff A, Kangas J, Coelho LP, Quinn S, Murphy RF: Automated image analysis for high-content screening and analysis. J Biomol Screen. 2010, 15 (7): 726-734. 10.1177/1087057110370894.View ArticleGoogle Scholar
- Carpenter AE: Extracting rich information from images. Methods Mol Biol. 2009, 486: 193-211. 10.1007/978-1-60327-545-3_14.View ArticleGoogle Scholar
- Abraham VC, Taylor DL, Haskins JR: High content screening applied to large-scale cell biology. Trends Biotechnol. 2004, 22 (1): 15-22. 10.1016/j.tibtech.2003.10.012.View ArticleGoogle Scholar
- Cervantes S, Prudhomme J, Carter D, Gopi KG, Li Q, Chang YT, Le Roch KG: High-content live cell imaging with RNA probes: advancements in high-throughput antimalarial drug discovery. BMC Cell Biol. 2009, 10: 45-10.1186/1471-2121-10-45.View ArticleGoogle Scholar
- Nath S, Spencer VA, Han J, Chang H, Zhang K, Fontenay GV, Anderson C, Hyman JM, Nilsen-Hamilton M, Chang YT, et al: Identification of fluorescent compounds with non-specific binding property via high throughput live cell microscopy. PLoS One. 2012, 7 (1): e28802-10.1371/journal.pone.0028802.View ArticleGoogle Scholar
- Lee JS, Kim YK, Kim HJ, Hajar S, Tan YL, Kang NY, Ng SH, Yoon CN, Chang YT: Identification of cancer cell-line origins using fluorescence image-based phenomic screening. PLoS One. 2012, 7 (2): e32096-10.1371/journal.pone.0032096.View ArticleGoogle Scholar
- Murphy RF, Boland MV, Velliste M: Towards a systematics for protein subcelluar location: quantitative description of protein localization patterns and automated analysis of fluorescence microscope images. Proc Int Conf Intell Syst Mol Biol. 2000, 8: 251-259.Google Scholar
- Newberg J, Hua J, Murphy RF: Location proteomics: systematic determination of protein subcellular location. Methods Mol Biol. 2009, 500: 313-332. 10.1007/978-1-59745-525-1_11.View ArticleGoogle Scholar
- Murphy RF: Cytomics and location proteomics: automated interpretation of subcellular patterns in fluorescence microscope images. Cytometry A. 2005, 67 (1): 1-3.View ArticleGoogle Scholar
- Shedden K, Brumer J, Chang YT, Rosania GR: Chemoinformatic analysis of a supertargeted combinatorial library of styryl molecules. J Chem Inf Comput Sci. 2003, 43 (6): 2068-2080. 10.1021/ci0341215.View ArticleGoogle Scholar
- Rosania GR, Lee JW, Ding L, Yoon HS, Chang YT: Combinatorial approach to organelle-targeted fluorescent library based on the styryl scaffold. J Am Chem Soc. 2003, 125 (5): 1130-1131. 10.1021/ja027587x.View ArticleGoogle Scholar
- Lee JW, Jung M, Rosania GR, Chang YT: Development of novel cell-permeable DNA sensitive dyes using combinatorial synthesis and cell-based screening. Chem Commun (Camb). 2003, 1852-1853. 15
- Shedden K, Li Q, Liu F, Chang YT, Rosania GR: Machine vision-assisted analysis of structure-localization relationships in a combinatorial library of prospective bioimaging probes. Cytometry A. 2009, 75 (6): 482-493.View ArticleGoogle Scholar
- Shedden K, Rosania GR: Chemical address tags of fluorescent bioimaging probes. Cytometry A. 2010, 77 (5): 429-438.Google Scholar
- Posada MM, Shedden K, Chang YT, Li Q, Rosania GR: Prospecting for live cell bioimaging probes with cheminformatic assisted image array (CAIA). Proc IEEE Int Symp Biomed Imaging. 2007, 1-3: 1108-1111. IEEE, New York, NYGoogle Scholar
- Zhang X, Shedden K, Rosania GR: A cell-based molecular transport simulator for pharmacokinetic prediction and cheminformatic exploration. Mol Pharm. 2006, 3 (6): 704-716. 10.1021/mp060046k.View ArticleGoogle Scholar
- Zheng N, Tsai HN, Zhang X, Shedden K, Rosania GR: The subcellular distribution of small molecules: a meta-analysis. Mol Pharm. 2011, 8 (5): 1611-1618. 10.1021/mp200093z.View ArticleGoogle Scholar
- Zheng N, Tsai HN, Zhang X, Rosania GR: The subcellular distribution of small molecules: from pharmacokinetics to synthetic biology. Mol Pharm. 2011, 8 (5): 1619-1628. 10.1021/mp200092v.View ArticleGoogle Scholar
- Trapp S, Rosania GR, Horobin RW, Kornhuber J: Quantitative modeling of selective lysosomal targeting for drug design. Eur Biophys J. 2008, 37 (8): 1317-1328. 10.1007/s00249-008-0338-4.View ArticleGoogle Scholar
- Zhang X, Zheng N, Rosania GR: Simulation-based cheminformatic analysis of organelle-targeted molecules: lysosomotropic monobasic amines. J Comput Aided Mol Des. 2008, 22 (9): 629-645. 10.1007/s10822-008-9194-7.View ArticleGoogle Scholar
- Yu JY, Zheng N, Mane G, Min KA, Hinestroza JP, Zhu H, Stringer KA, Rosania GR: A cell-based computational modeling approach for developing site-directed molecular probes. PLoS Comput Biol. 2012, 8 (2): e1002378-10.1371/journal.pcbi.1002378.View ArticleGoogle Scholar
- Yu JY, Rosania GR: Cell-based multiscale computational modeling of small molecule absorption and retention in the lungs. Pharm Res. 2010, 27 (3): 457-467. 10.1007/s11095-009-0034-9.View ArticleGoogle Scholar
- Zhang X, Zheng N, Zou P, Zhu H, Hinestroza JP, Rosania GR: Cells on pores: a simulation-driven analysis of transcellular small molecule transport. Mol Pharm. 2010, 7 (2): 456-467. 10.1021/mp9001969.View ArticleGoogle Scholar
- Trapp S, Horobin RW: A predictive model for the selective accumulation of chemicals in tumor cells. Eur Biophys J Biophys Lett. 2005, 34 (7): 959-966. 10.1007/s00249-005-0472-1.View ArticleGoogle Scholar
- Trapp S, Horobin RW: A predictive model for the selective accumulation of chemicals in tumor cells. Eur Biophys J. 2005, 34 (7): 959-966. 10.1007/s00249-005-0472-1.View ArticleGoogle Scholar
- Lang AS, Bradley JC: Chemistry in second life. Chem Cent J. 2009, 3: 14-10.1186/1752-153X-3-14.View ArticleGoogle Scholar
- Boulos MN, Hetherington L, Wheeler S: Second Life: an overview of the potential of 3-D virtual worlds in medical and health education. Health Info Libr J. 2007, 24 (4): 233-245. 10.1111/j.1471-1842.2007.00733.x.View ArticleGoogle Scholar
- Bradley JC, Lancashire RJ, Lang AS, Williams AJ: The Spectral Game: leveraging Open Data and crowdsourcing for education. J Cheminform. 2009, 1 (1): 9-10.1186/1758-2946-1-9.View ArticleGoogle Scholar
- Patel V, Lee H, Taylor D, Aggarwal R, Kinross J, Darzi A: Virtual worlds are an innovative tool for medical device training in a simulated environment. Stud Health Technol Inform. 2012, 173: 338-343.Google Scholar
- Fantin VR, Leder P: F16, a mitochondriotoxic compound, triggers apoptosis or necrosis depending on the genetic background of the target carcinoma cell. Cancer Res. 2004, 64 (1): 329-336. 10.1158/0008-5472.CAN-03-0899.View ArticleGoogle Scholar
- Fantin VR, Leder P: Mitochondriotoxic compounds for cancer therapy. Oncogene. 2006, 25 (34): 4787-4797. 10.1038/sj.onc.1209599.View ArticleGoogle Scholar
- Fantin VR, Berardi MJ, Scorrano L, Korsmeyer SJ, Leder P: A novel mitochondriotoxic small molecule that selectively inhibits tumor cell growth. Cancer Cell. 2002, 2 (1): 29-42. 10.1016/S1535-6108(02)00082-X.View ArticleGoogle Scholar
- Shedden K, Xie XT, Chandaroy P, Chang YT, Rosania GR: Expulsion of small molecules in vesicles shed by cancer cells: association with gene expression and chemosensitivity profiles. Cancer Res. 2003, 63 (15): 4331-4337.Google Scholar
- Li Q, Lee JS, Ha C, Park CB, Yang G, Gan WB, Chang YT: Solid-phase synthesis of styryl dyes and their application as amyloid sensors. Angew Chem Int Ed Engl. 2004, 43 (46): 6331-6335. 10.1002/anie.200461600.View ArticleGoogle Scholar
- Li Q, Min J, Ahn YH, Namm J, Kim EM, Lui R, Kim HY, Ji Y, Wu H, Wisniewski T, et al: Styryl-based compounds as potential in vivo imaging agents for beta-amyloid plaques. Chembiochem. 2007, 8 (14): 1679-1687. 10.1002/cbic.200700154.View ArticleGoogle Scholar
- Loo LH, Wu LF, Altschuler SJ: Image-based multivariate profiling of drug responses from single cells. Nat Methods. 2007, 4 (5): 445-453.Google Scholar
- Perlman ZE, Slack MD, Feng Y, Mitchison TJ, Wu LF, Altschuler SJ: Multidimensional drug profiling by automated microscopy. Science. 2004, 306 (5699): 1194-1198. 10.1126/science.1100709.View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.