PanHunter FAQ page

• The major part of Evotec’s revenue generates from CRO-services that focus on the partnered drug discovery
• 14 different sites where the scientists are working on various projects: diabetes and its complications, inflammation and immunology, iPSC-based research and so on
• Many studies are based on high-throughput compound screening and patient materials

• Transcriptomics > genomics = proteomics > clinical data > metabolomics

• We have a bioinformatics team who develop internal solutions along the needs of individual research projects and time spans

• Bulk RNA-seq: deep sequencing, usually paired-end
• Single cell RNA-seq, 3’-focused
• Spacial transcriptomics, 3’-focused
• Screen-seq / Plate-seq: 384-well plate, 3’-focused

• Unbiased experiment for large-scale drug / compound screens (384-well plate)
• Automated robust workflow, every sample is handled with equal condition

• PanHunter is truly multi-omics as it has means of integrating more than one omics type and supports multi-omics analysis. Compared to other software it is relatively fast and offers most comprehensive set of tools needed for omics analysis. It also comes with bioinfo support and is built in a modular way, which makes it very easy to add additional app if needed.

• When having omics data (transcriptomics, proteomics, genomics, metabolomics…)
• When obtaining big and complex data that can be converted to count matrices (phenotypic readout, cell painting, flow cytometry…)
• When robust pipeline and algorithms are required
• When collaborative work space is beneficial

• When the scientists only perform simple bioassays such as western-blot, qPCR, ELISA, fluorescence read-out from a couple of antibodies or reagents etc.
• When experiment protocol (e.g. cell differentiation protocol) is short: only few steps and very time-efficient. In this case, conducting omics studies (~3 weeks to obtain transcriptomics data) would be too long and expensive, and it’s not possible to keep some cells viable for so long.

• Patient Data app is especially useful for post experiment stratification.

• Mixture of peer reviewed, internal development and external sources.
• The information is available in PanHunter documentation, which can be accessed by clicking on the Help button in PanHunter.

• No, there are no tools which can be used for text and picture data mining.

PanHunter is a web-based analysis platform that runs on Evotec servers. The users can open one of the below URLs with a modern web browser of choice, to log in and start PanHunter.

• https://panhunter.com/producttour/

• The username and the password are the same as the Microsoft Windows ones.
• In case you are not based in Germany and/or cannot find your PanHunter projects after logging in, please try to log in with username@corp instead of your Windows username alone.

• Google Chrome is recommended; nevertheless Mozilla Firefox and Microsoft Edge should work just fine.
• If you have any problems accessing PanHunter via a stably working VPN, please try to use an incognito window.

• Yes, PanHunter supports collaborative access. It makes sure one does not overwrite other people’s inputs; everything is saved on the fly in a dedicated secure environment.

• All accessible PanHunter projects of the user are shown as separate tabs at the top of the Start page.
• Within a specific PanHunter project, users can find its related studies and access them via the included available apps.

• Please send the request to panhunter_support@evotec.com, include the project lead in cc and provide the following information:

1. Windows username of the user
2. Name of the PanHunter project

• Requests can be handed in through SharePoint EvoFlow, Admin app or email to panhunter_support@evotec.com
• Required information:
  1. Project name: desired name to be displayed in PanHunter
  2. Project ID: internal unique identifier of the project, only lower-case letters and/or numbers
  3. Species
  4. Name of the project lead(s)
  5. Project billing code, i.e. IFS code
  6. Optional: short description of the project

• For transcriptomics datasets, PanHunter accepts and prefers fastq files generated directly from sequencing machines.

• For other omics types or assays such as proteomics, metabolomics, phenotypic readouts, cell painting, etc., count matrices, bam files, R- and python-generated files can be integrated smoothly.

• Yes, PanHunter can support it. Human genome contains approximately 120,000 genes.

• Yes, it is supported.

Organisms listed below are fully supported. Integration of other species or customed metadata is possible; discussions on the requirements at an early stage is needed.

• Human (Homo sapiens/Hs)
• Rat (Rattus norvegicus/Rn)
• Mouse (Mus musculus/Mm)
• Macaques (Macaca mulatta/Mmul, Macaca fascicularis/Mf)
• Pig (Sus scrofa/Ss)
• Chinese hamster (Cricetulus griseus/Cgri)
• Dog (Canis lupus familiaris/Cluf)

• It depends primarily on the provided data; fast and robust procedure has been established for fastq files. Other data types can be integrated within a workweek.

• This feature is under development. At the moment, it is not visible via the graphical user interface.

Yes, fastq, bam and count matrices can be provided. However, files that are generated during pre-processing (e.g., demultiplexing) are usually discarded due to the fact that they are uninformative and space-intensive.

Following data files can be exported directly from the New Comparisons app in PanHunter:

• Transcriptomcis data:

1. raw: raw counts per gene
2. norm: count per million (CPM)
3. lognorm: log2 of the CPM values

• For ScreenSeq data:

The formula for calculating log-normalization of ScreenSeq data is $log2(CPM/100 + 1)$. CPM is divided by 100 to adapt to initially (now substantially improved) shallow gene counts where counts per 10K count unit was more applicable. A pseudocount is added to avoid 0 count that is not accepted by log function.

• Proteomics data:

1. norm: normalized protein intensities
2. lognorm: log2 of normalized protein intensities

• It is feasible, however not via the graphical user interface yet. Please contact us through the Contact Support button or panhunter_support@evotec.com.

Selecting samples is performed as follows:

1. In the left panel, there is a Sample selection menu. By default, it is empty and the user needs to select a study first.
2. To select a study, please click on the empty Values field in Filter: Study section and choose at least one study from the drop down menu.
3. Additional specifications can be defined by selecting different modifiers from the drop down menu at the bottom of the Sample selection. Available modifiers are Filter categ., Filter num., Join Columns, Column binning and Enrich .
4. After selecting the desired modifier and clicking on +, new sections should appear accordingly.
5. For detailed explanations, please check out the tutorial video: How to use the New Sample Selector?

Enabling custom annotations is performed as follows:

1. The modifier Enrich needs to be added to the Sample selection menu.
2. After Enrich section has appeared, Custom annotations should be selected as the data source for the enrichment.
3. The Sample annotations section should be available for editing afterwards, in the left panel.

Samples dots can be selected via:

1. Lasso tool: by clicking and holding the right mouse button, and circling the samples of interest
2. Figure legend: by clicking on the legend item, the samples in the corresponding category will be selected

The users can combine multiple selections by holding the shift key while selecting them. Holding the ctrl key, on the other hand, can deselect the samples.

• Transcriptomics data; bulk RNA-seq, single-cell RNA-seq and plate RNA-seq/ScreenSeq are applicable.

• For proteomics data, please use the Proteomics QC app.

• With advanced technology nowadays, >80 % uniquely mapped reads can be obtained from a sample with good quality. Additionally, >90% demultiplexed reads can be achieved by ScreenSeq. However the definition of a good sample strongly depends on the experimental setup (e.g. model organism or a cell line).

• The mentioned two indicators: uniquely.mapped.reads.percent and PercentDemux, can be found in the Alignment stats and in the Files subtab of plateRNA-Seq, respectively.

• Low quality does not come without a reason and we never cannot exclude that the same reason would affect the gene expression profile. Moreover, low quality typically leads to an effectivly shallow quantification of gene expression, which leverages the contribution of noise.
• If the low quality samples are kept, it may (and most probably will) affect the interpretation of gene expression quantification and differential expression analysis.

• During plate RNA-seq, each well of the plate is labeled with a specific barcode. To check the behavior of the barcodes, a visual overview is provided in the Plate/Library subtab in the plateRNA-Seq section. Detailed information of selected barcodes can be inspected in the Barcode Corrections subtab.

• Barcode bias describes the fact that some barcodes generate less reads, i.e. lower sequencing and demultiplexing efficiencies, than others due to its own nature. Nevertheless, all reads are corrected at the end in PanHunter.

• When one mismatch occurs during synthesis or sequencing, it’s automatically corrected at the barcode correction step during data pre-processing.

• Yes, only the detected genes are shown in PanHunter.

• Outliers are usually flagged by our application experts already while performing the raw data QC during data integration. They are excluded since they can negatively influence the analysis results.
• If desired, the users can display and include the excluded samples in analyses through the PanHunter graphical interface, by activating the Advanced Settings via the slider and selecting Show excluded samples in Sample selection section.

• Yes, depending on the research question, one can compare between different samples, studies, omics types, and even different species.

• Please contact PanHunter support (panhunter_support@evotec.com) for transferring selected studies to desired projects or combining projects.

Yes, multiple apps in PanHunter support side-by-side comparison and analysis between top tables calculated with different omics types.

• Top Tables: Results of up to 4 top tables can be compared side-by-side in the Compare Top Tables tab.
• Cross-Comparison: Samples from different datasets can be linked and visualized in dimension reduction plots.
• MA Plot: Top Tables calculated using different datasets can be visualized in volcano plots.
• Pathway Mapping: Top Tables calculated using different datasets can be visualized on pathway schemes.

• Yes, Patient Data app is dedicated for exploring clinical data and correlating it with omics results. It´s also possible to enrich patient data information via the Enrich modifier in Sample selection and use it in other apps. Please see Sample Selection for more details.

• A PCA plot, or a Principal Component Analysis plot, is a dimension reduction plot where samples are clustered together based on their similarity while preserving the maximum amount of information. PCA Component 1 and PCA Component 2 are displayed on plot axes automatically as components with highest cluster forming significance. That is, the closer the sample dots cluster together in a PCA plot, the similar their transcriptome/proteome profiles are.

• In PanHunter, PCA plot is accompanied by additional bar plot where all PCA Components are displayed with corresponding standard deviation.

Behind the scenes, a series of steps are done before the samples are plotted in PCA space.

• By default, only the top 500 features with the most variance are included in the dataset.
• Data are loaded from the matrices with the respective main datatype given its platform.
• NA values are mean imputed
• Values are centered but not scaled
• The calculation is done with done either with irlba::prcomp_irlba or, if this fails, with stats::prcomp

Minor numerical differences can lead to the inversion of Principal Component (PC) transformed values. As a result, visualizations of the data may appear mirrored along the respective axis of the transformed dimension. However, this does not affect the overall interpretation of PCA results.

• Algorithm based on which samples are clustered calculates pairwise distances. We use Mann-Whitney U test on intra- and inter-class distances (for categorical) and Spearman‘s rank correlations between pairwise distances and differences of a variable (for numerical).

• When choosing Comparison analysis as the Data for dimension reduction, a coverage in % will be provided together with the available Comparison Groups. It is the proportion of comparisons by the number of SampleIDs in the Study.
• For example, if there are 4 compounds and 1 control, each having the same number of samples, the coverage will be 4/5 = 80%.

• The categorical and numerical variables in the sample metadata as well as saved custom annotations can be used to specify the comparison parameters for running differential analysis in the New Comparison app.
• In case of confounding/batch effects, it is recommended to include associated covariates in the model, if it keeps model complexity reasonable for model fitting. For example, PlateID is included as covariate for ScreenSeq data.

Two statistical methods are available in PanHunter:

1. DESeq2: the method is optimal for data containing discrete numbers, e.g. count values from transcriptomics data sets.
2. Limma: the method is optimal for data containing continuous numbers such as feature intensities from proteomics and metabolomics data sets.

By default, the optimal statistical method will be automatically selected by PanHunter. Nevertheless, if desire, another method can be selected manually.

• Limma is less compute intensive and can generate results faster for larger studies. Nevertheless, users can apply DESeq2 manually if needed.

• Yes, if we perform differential expression analysis with Limma for RNA-Seq data, Voom precision coefficients are applied.

• Indeed ScreenSeq studies usually have >100 samples, however, DESeq2 is recommended especially when one of the groups in a comparison contains a small number of samples. The suggestion is based on our internal benchmarking studies as well as the fact that the statistical model used in DESeq2, negative binomial distribution, is more consistent in comparison with the “continuous” statistical model used in Limma. Consistency needs to be considered as there is a limited sequencing depth in ScreenSeq for individual samples, counts for genes are pronouncedly discrete, especially for lower expressed one.

• Currently there is no option to see the MA/Volcano plot only for one previously saved top table. Our developers are aware of this and are working on a solution. It is possible, however, to select the same top table in both entries in the MA Plot app. In this case, both the correlation plots between log2 fold change and between abundance of the two datasets should display perfect matches as the input top tables are identical.

• Gene Info app is dedicated for checking the expression values of a single gene across different conditions. Please specify the search parameters first, such as the studies of interest (i.e. cancer types in this case) and the target gene or feature, then the results will be present in a tabular or graphic view, in the Results or Graphic tab, respectively.

• Yes, all tables in PanHunter can be customized. Users can add and remove any column via the table setting shown as a three-bar icon located on the upper left corner of a table.

• Yes, they are updated regularly.

• No, currently it´s not allowed to create your own pathways as this could result in various problems when operating PanHunter. Current available pathways are from Wikipathways.

• Yes, visualizing fold changes from two top tables calculated from different omics types on the same pathway is possible, as long as the data are stored in the same project. Transcriptomics, proteomics and metabolomics data can be applied here (although the coverage of the mapped metabolites is low due to the fact that there is no comprehensive metabolite database yet).

• The Signature Visualization app provides an interactive visualization of a signature against the top table genes. The bar represents genes from the selected top table: down-regulated gene in orange and up-regulated genes in blue; ordered according to the fold-change. The dots above the bar represent the signature genes; their locations correspond to the ordered top table genes in the bar and their colors reflect to the gene regulation of the signature.

• Different gene sets are compared accoding to the Jaccard index. Similar terms obtained from selected databases (ChIP Atlas, GO, MSigDB and Wikipathways) are clustered and visualized in the 2D plot.

• All genes that are associated with the respective gene set are shown, as long as no additional filtering is applied.
• There is currently not possible to only show the significantly regulated ones within PanHunter. When displaying FDRs of GO terms, this method is used for testing and Fisher/Wilcoxon/KS are used for Wikipathways. However there is currently no method to subset to GO terms or wikipathways of interest.

• Single-cell RNA-seq requires at least 50,000 cells (1 million is recommended) as an input; 500 to 10,000 cells per sample should be analyzed; 100,000 reads per cell is ideal to maximize the identification of transcripts.

• Commonly, microdroplet-based methods are applied during single-cell RNA-seq; i.e. individual cells are encapsulated in a microdroplet, where the reverse transcription reaction takes place, converting RNAs to cDNAs. Ideally, each droplet should contain only one cell; however, there are situations which no or multiple cells are present in one droplet.

• When a droplet does not contain a cell, only the ambient RNAs are sequenced which produces background noise. When a droplet contains more than one cell, called multiplet, high number of expressed genes in comparison to the average level would be detected. This can be identified and corrected during pre-processing and data integration.

• Moreover, a droplet can contain one or multiple dying cells. In this case, high numbers of mitochondrial transcripts will be detected. One should note that the expression level of mitochondrial genes depends highly on the tissue of interest. For instance, muscle, heart, liver, kidney, and to a certain extent, brain tissues are considered as mitochondria-rich tissues.

• The full name of CITE-seq is Cellular Indexing of Transcriptomes and Epitopes by Sequencing. It is a method to perform single cell transcriptomics along with gaining quantitative and qualitative information of surface protein markers using available antibodies.

• One barcode corresponds to one cell; namely, one barcode comes from one cell and one cell is labeled with one barcode (in best case, when no multiplet present). However, one UMI corresponds to one transcript; namely, one UMI comes from one transcript generated by a cell but one cell contains multiple transcripts and therefore is labeled by multiple UMIs.

• Technically, UMIs are added before PCR amplification in order to reduce errors and quantitative bias introduced during the amplification. A length of 16 nt barcode and 10 nt UMI are commonly used in the 10x genomics workflow for scRNA-seq.

• Yes, PanHunter provides both predicted classification based on curated database and the possibility to annatate the cells manually.

• Yes, it is possible to generate sub-clusters using the scRNA-Seq Browser. Please select the cells of interest by holding the right mouse button and circling the target cells in the Overview tab. This will save its coordinates automatically and can be further re-calculated, visualized and inspected in the Sub-clustering tab.

• Yes, pseudotime analysis (we currently use Slingshot) can be done by in the Pseudotime tab. All lineages are available in the resulting dropdown menu and two lineages can be selected to visualize side-by-side. Differential expression calculation on top of pseudotime lineages can be also calculated.

• Uploading the data files to client’s cloud-based storage is preferred. We are well-expericenced with AWS S3 buckets and sftp.

• In case a storage bucket is not available, we can set up sftp server on our end to transfer the data. Other solutions need to be discussed.

• It is possible, however requires highly complicated process. Case-by-case negotiations at business and legal levels between Evotec and the client are required.

• The number of integrated databases in PanHunter has been increasing; below are some examples of the integrated ones:

Gene Ontology, WikiPathways, ChIP-Atlas, UniProt, Pfam, Creeds, The human protein atlas, Protein data bank, OMIM, ChEMBL, DrugMAtrix, MSigDB, BioGRID, ConnectivityMap, e!Ensembl, TARGET, GTExPortal, NCBI, NIH

• No, KEGG is not included. It is a commercial database which the licencing contracts and the pathway updates are difficult to navigate.

• Yes, ConnectivityMap (CMAP) signatures are included in PanHunter. The user can use them, e.g., in the Signature Visualization app to compare a Top Table of interest against the signatures from CMAP database.

• The Chem Info app is dedicated to omics compounds studies. The chemical information are part of this study. It can be either publicly available compounds or Evotec compounds.

• More than 80 publicly available data sets including transcriptomics, proteomics, single cell and clinical data are included in following projects: TCGA, Public Cell Atlas, Rat Body Map, Cancer Cell Atlas, ScreenSeq Demo and Proteomics Demo. These projects can be openly accessible.

• Publicly available data sets from TCGA, Human Cell Atlas etc. are accessible for every user. Additional public knowledge can be integrated by request.

• Evotec molecular patient database (E.MPD) includes molecular and clinical data of several disease areas from different patient cohorts. Access to the proprietary databases needs to be discussed.

As a first approach, you can try to reset the user settings (see below) of the app that is causing problems. If this does not help, or if the same problem shows up repeatedly, please reach out to the PanHunter support in one of the following ways:

• Through Contact support button - fill in the form and provide all the details
• Email to the panhunter_support@evotec.com with a description of your problem, the name of the app that is not working, and the project you are using. If possible, add a snapshot link (created by clicking on “Snapshot” in the upper-right corner on your screen) or a screenshot to the e-mail and steps to reproduce the error.

• Resetting user setting for an app is always a good idea if you are stuck within one app.
• To do this, there are two options:
  • Hover the mouse over an app and click on the little gear icon showing on its upper-right corner. This will open up an app-specific configuration pop-up. Here, you can click the “Reset user settings” button to reset the settings for the selected app.
  • If there is an “Admin” link in the lower right corner of the PanHunter start page, click it to enter the Admin page. Here, go to the “User settings” tab, select the app for which you would like to reset your settings, and press “Reset user settings”. Now, when restarting the app, you should see a clean interface. CAUTION: Please note that resetting your user settings will deselect all samples and remove all parameter values in that app. This might be undesirable and thus you should try this fix only if you are aware of the consequences.

• Please check if Outlook has been set as the default app for email in your Windows settings. If not, go to the Windows Settings -> Apps -> Default Apps and choose Outlook as the default app for the E-mail
• If setting Outlook as the default app doesn’t resolve the problem, please contact us via panhunter_support@evotec.com

• We have prepared a series of training videos that covers most of the apps, their functionality, and how to use them. We also discuss some best practices within those videos.

• If you, or a member of your team, is new to PanHunter or would like to receive additional training (exceeding what is covered by the training videos), send an e-mail to panhunter_support@evotec.com and ask for a respective training session.

• While working with PanHunter, you will see the help link at the upper-right corner of the page. This leads you to the documentation of PanHunter.

• In case PanHunter Help documentation does not provide answer or solution to your issue, please use the Contact Support buttom or write an email to panhunter_support@evotec.com and our team will try our best to help you.