Booklet for the PiER

Hai Fang

National Research Center for Translational Medicine at Shanghai, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China

2022-05-01

Section 1 Background

FIGURE 1.1: The logo for the PiER. The above-water pillar structure in red (symbolising the infrastructure) and water waves in blue (by analogy the piano stave) collectively illustrate the web-based PiER facilities enabling ab initio and real-time genetic target prioritisation.

---

Motivation

The field of target discovery has been advanced by genetics-led target prioritisation approaches. Integrative prioritisation for early-stage genetic target discovery has proven cost-effective in promoting the translational use of disease genetic associations, which is increasingly recognised in reducing drug attrition rate in late-stage clinical trials.

Design

Building on the verified Pi approach (see Nature Genetics 2019), here I introduce web-based servers/facilities called PiER. The PiER is free and open to all users and there is no login requirement, allowing the users to perform ab initio and real-time target prioritisation harnessing human disease genetics, functional genomics and protein interactions.

By analogy to the piano stave, the PiER consists of five horizontal lines, with three lines representing the elementary facility (eV2CG, eCG2PG and eCrosstalk), each doing specific tasks on their own, and the rest two lines signifying the combinatory facility (cTGene and cTCrosstalk).

• eV2CG, linking variants to core genes; see Example Output

• eCG2PG, networking core genes to peripheral genes; see Example Output

• eCrosstalk, identifying the crosstalk between pathways; see Example Output

• cTGene, prioritising targets at the gene level; see Example Output

• cTCrosstalk, prioritising targets at the crosstalk level; see Example Output

Section 2 Facilities

The elementary facility supports three specific tasks, including three online tools: (i) eV2CG, utilising functional genomics to link disease-associated variants (including those located at the non-coding genome) to core genes likely responsible for genetic associations; (ii) eCG2PG, using knowledge of protein interactions to ‘network’ core genes with each other and with additional peripheral genes as well, producing a ranked list of core and peripheral genes; and (iii) eCrosstalk, exploiting the information of pathway-derived interactions to identify highly ranked genes that mediate the crosstalk between molecular pathways. By chaining together elementary tasks supported in the elementary facility, the combinatory facility enables the automation of genetics-led and network-based integrative prioritisation for genetic targets, both at the gene level (cTGene) and at the crosstalk level (cTCrosstalk). Notably, in addition to target crosstalk, the cTCrosstalk further supports target pathway prioritisation and crosstalk-based drug repurposing analysis (that is, repositioning approved drugs from original disease indications into new ones).

---

FIGURE 2.1: Schematic illustration of two facilities supported in the PiER.

Section 3 Compatibility

TABLE 3.1: A summary of the PiER browser compatibility.

	MacOS (Big Sur)	Windows (10)	Linux (Ubuntu)
Safari	14.1.2	N/A	N/A
Microsoft Edge	N/A	85.0.564.67	N/A
Google Chrome	96.0.4664.110	90.0.4430.93	96.0.4664.110
Firefox	95.0.2	95.0.2	95.0.2

Section 4 Runtime

TABLE 4.1: A summary of the runtime (on the server and client sides) per tool estimated using Google Chrome.

Facilities	Tools	Runtime (Server + Client)
Elementary	eV2CG	(67 + 82) seconds
Elementary	eCG2PG	(15 + 70) seconds
Elementary	eCrosstalk	(53 + 71) seconds
Combinatory	cTGene	(90 + 91) seconds
Combinatory	cTCrosstalk	(143 + 97) seconds

Section 5 Frontpage

FIGURE 5.1: The landing frontpage (visited using Google Chrome in MacBook Pro) of the PiER, featuring two facilities (elementary and combinatory). The elementary facility includes: (i) eV2CG, linking disease associated variants (particularly located at the non-coding genomic region) to core genes likely responsible for genetic associations, based on either promoter capture Hi-C (PCHi-C, that is, conformation evidence), quantitative trait locus (QTL) mapping (that is, genetic regulation of gene expression or protein abundance), or simply genomic proximity; (ii) eCG2PG, using knowledge of protein interactions to ‘network’ core genes with each other and with additional peripheral genes as well, producing a ranked list of core and peripheral genes; and (iii) eCrosstalk, exploiting the information of pathway-derived interactions to identify highly-ranked genes that mediate the crosstalk between molecular pathways. By chaining together elementary tasks supported in the elementary facility, the combinatory facility enables automation of genetics-led and network-based integrative prioritisation for genetic targets: (iv) at the gene level (cTGene); and (v) at the crosstalk level (cTCrosstalk). Also included is the tutorial-like booklet (in an HTML format) describing step-by-step instructions on how to use.

Section 6 Development

The PiER was developed using a next-generation Perl web framework Mojolicious that requires nearly zero-effort maintenance for interface updates. The PiER was also built using Bootstrap that supports the mobile-first and responsive webserver. The source codes are made available at GitHub.

---

FIGURE 6.1: The screenshots for the PiER visited using Google Chrome in iPhone. Left: the frontpage; Right: the eV2CG interface.

Section 7 Help buttons

Each user-request interface has the Show/Hide Info toggle button that contains the help information on use, including the details on input, output, mechanism and other useful information, while the Example I/O button showcases the example input/output. For example, shown below is the screenshot in the cTCrosstalk interface.

---

FIGURE 7.1: The screenshots for the Show/Hide Info toggle button in the cTCrosstalk interface.

Section 8 Error messages

The error messages will be displayed, for example, if the input into the cTCrosstalk is invalid (see the screenshot below). Notably, in the results page, a summary of input data is also returned to the users for the reference.

---

FIGURE 8.1: The screenshot for the error messages shown when the input is invalid, for example, in the cTCrosstalk interface.

Section 9 eV2CG

9.1 Interface

Input

• Step 1: a list of user-input SNPs, with 1st column for dbSNP rsIDs and 2nd column for significance info (p-values between 0 and 1). The error message will be displayed if the input is invalid. Example input data are shared genetic variants identified from cross-disease genome-wide association studies in inflammatory disorders; see Nature Genetics 2016.

Mechanism

• Step 2: includes SNPs in Linkage Disequilibrium (LD). By default, input SNPs with a typical threshold (p-value < 5e−8) are considered, and additional SNPs in linkage disequilibrium (R2 < 0.8) can be also included according to the European population.

• Step 3: uses genomic proximity, quantitative trait locus (QTL), or promoter capture Hi-C (PCHi-C) to identify core genes.

• More Controls: fine-tunes parameters involved in steps described above.

Output

• Example Output includes two interactive tables for core genes and evidence used, and a manhattan plot (illustrating scored core genes color-coded by chromosomes). A summary of input data and the runtime (computed on the server side) is also returned to the users for the reference.

---

FIGURE 9.1: The interface of eV2CG, linking disease associated variants (particularly located at the non-coding genomic region) to (core) genes likely responsible for associations, based on either promoter capture Hi-C (PCHi-C; conformation evidence), quantitative trait locus (QTL) mapping (that is, genetic regulation of gene expression or protein abundance), or simply genomic proximity. The Show/Hide Info toggle button contains the help information on how to use the eV2CG, including input, output, mechanism, etc.

9.2 Linking results

• Under the tab Output: core genes, Manhattan plot illustrates scored core genes that are color-coded by chromosomes. Also provided is the downloadable PDF file.

• Under the tab Output: core genes, An interactive table lists core genes linked from the input SNPs, with scores quantifying the level of genes responsible for genetic associations (capped at 100). Genes are cross-referenced and hyperlinked to GeneCards. Also provided is the column Evidence used to define core genes.

• Under the tab Output: core genes, Evidence table for core genes, showing which SNPs (see the column SNPs) are used to define core genes (the column Core genes) based on which evidence (see the column Evidence). The column SNP type tells the SNP type (either Input for use-input SNPs or LD for LD SNPs). Notably, the column Evidence details datasets used: the prefix Proximity_ indicative of SNPs in the proximity, the prefix PCHiC_ for PCHi-C datasets, and the prefix QTL_ for e/pQTL datasets.

---

FIGURE 9.2: Interactive results for the eV2CG. Under the tab Output: core genes is a manhattan plot illustrating scores for core genes. The user-input data under the tab Input into eV2CG are also returned for the exploration.

---

FIGURE 9.3: Two tabular displays about core genes (top) and evidence (bottom) under the tab Output: core genes.

Section 10 eCG2PG

10.1 Interface

Input

• Step 1: a list of user-defined core genes, with 1st column for gene symbols, 2nd columns for weights (positive values), such as results from eV2CG above. The error message will be displayed if the input is invalid.

Mechanism

• Step 2: networks core genes with each other and with additional (peripheral) genes based on the knowledge of protein interactions, generating a ranked list of core and peripheral genes. It is achieved using the random walk with restart (RWW) algorithm. By default, the restarting probability of 0.7 is set, empirically optimised for immune-mediated diseases; selecting a value smaller than 0.6 is not recommended as there is a higher chance to expect low performance.

• More Controls: fine-tunes parameters involved in steps described above.

Output

• Example Output includes an interactive table for core and peripheral genes, and a manhattan plot (illustrating scores for genes color-coded by chromosomes). A summary of input data and the runtime (computed on the server side) is also returned to the users for the reference.

---

FIGURE 10.1: The interface of the eCG2PG, using the knowledge of protein interactions to ‘network’ core genes with each other and with additional (peripheral) genes as well, generating a ranked list of core and peripheral genes. The Show/Hide Info toggle button contains the help information on how to use the eCG2PG, including input, output, mechanism, etc.

10.2 Networking results

• Under the tab Output: core and peripheral genes, Manhattan plot illustrates affinity scores for genes that are color-coded by chromosomes. Also provided is the downloadable PDF file.

• Under the tab Output: core and peripheral genes, An interactive table lists core and peripheral genes, with scores quantifying the affinity to core genes (sum up to 1). Genes are cross-referenced and hyperlinked to GeneCards.

---

FIGURE 10.2: Interactive results for the eCG2PG under the tab Output: core and peripheral genes. The user-input data the tab Input into eCG2PG are also returned for the exploration.

Section 11 eCrosstalk

11.1 Interface

Input

• Step 1: a ranked list of genes, with 1st column for gene symbols, 2nd columns for scores (positive values), such as results from eCG2PG above. The error message will be displayed if the input is invalid.

Mechanism

• Step 2: identifies the subnetwork of highly-ranked genes that mediate the crosstalk between molecular pathways. The significance (p-value) of observing the identified crosstalk by chance is estimated by a degree-preserving node permutation test.

Output

• Example Output includes an interactive table for pathway crosstalk genes, and a network visualisation (illustrating the crosstalk between pathways).

---

FIGURE 11.1: The interface of the eCrosstalk, exploiting the information of well-curated pathway-derived interactions to identify the subnetwork of highly ranked genes that mediate pathway crosstalk. The Show/Hide Info toggle button introducing how to use the eCrosstalk, including input, output, mechanism, etc.

11.2 Crosstalk results

• Under the tab Output: pathway crosstalk, A network visualisation illustrates crosstalk genes color-coded by input scores. The significance (p-value) of observing the identified crosstalk by chance is estimated by a degree-preserving node permutation test. Also provided is the downloadable PDF file.

• Under the tab Output: pathway crosstalk, An interactive table: lists crosstalk genes together with input scores. Genes are cross-referenced and hyperlinked to GeneCards.

---

FIGURE 11.2: Interactive results for the eCrosstalk under the tab Output: pathway crosstalk. The user-input data under the tab Input into eCrosstalk are also returned for the exploration.

Section 12 cTGene

12.1 Interface

Input

• Step 1: a list of user-input SNPs, with 1st column for dbSNP rsIDs and 2nd column for significance info (p-values between 0 and 1). The error message will be displayed if the input is invalid. Example input data are shared genetic variants identified from cross-disease genome-wide association studies in inflammatory disorders; see Nature Genetics 2016.

Mechanism

• Step 2: includes SNPs in Linkage Disequilibrium (LD). By default, input SNPs with a typical threshold (p-value < 5e−8) are considered, and additional SNPs in linkage disequilibrium (R2 < 0.8) can be also included according to the European population.

• Step 3: uses functional genomic datasets, including genomic proximity, quantitative trait locus (QTL) and promoter capture Hi-C (PCHi-C), to identify core genes.

• Step 4: networks core genes with each other and with additional (peripheral) genes based on the knowledge of protein interactions, generating a ranked list of core and peripheral genes. It is achieved using the random walk with restart (RWW) algorithm. By default, the restarting probability of 0.7 is set, empirically optimised for immune-mediated diseases; selecting a value smaller than 0.6 is not recommended as there is a higher chance to expect low performance.

• More Controls: fine-tunes parameters involved in steps described above.

Output

• Example Output includes a manhattan plot (illustrating priority rating for target genes color-coded by chromosomes), and two tabular displays about prioritisation and evidence. A summary of input data and the runtime (computed on the server side) is also returned to the users for the reference.

---

FIGURE 12.1: The interface of the cTGene, enabling/automating genetics-led and network-based identification and prioritisation of drug targets at the gene level. The Show/Hide Info toggle button contains the help information on how to use the cTGene, including input, output, mechanism, etc.

12.2 Prioritisation results

• Under the tab Output: target genes, Manhattan plot illustrates priority rating for target genes that are color-coded by chromosomes. Also provided is the downloadable PDF file.

• Under the tab Output: target genes, Prioritisation table lists all prioritised genes, each receiving 5-star priority rating (scored 0-5). Genes are cross-referenced and hyperlinked to GeneCards. The column Type tells the target gene type (either Core for core genes or Peripheral for peripheral genes). Also provided is a summary of evidence used to define core genes, including columns Proximity (evidence of genomic proximity), QTL (e/pQTL evidence) and PCHiC (conformation evidence).

• Under the tab Output: target genes, Evidence table for core genes, showing which SNPs (see the column SNPs) are used to define core genes (the column Core genes) based on which evidence (see the column Evidence). The column SNP type tells the SNP type (either Input for use-input SNPs or LD for LD SNPs). Notably, the column Evidence details datasets used: the prefix Proximity_ indicative of SNPs in the proximity, the prefix PCHiC_ for PCHi-C datasets, and the prefix QTL_ for e/pQTL datasets.

---

FIGURE 12.2: Prioritisation results for the cTGene. Under the tab Output: target genes is a manhattan plot illustrating priority rating for target genes. The user-input data under the tab Input into cTGene are also returned for the exploration.

---

FIGURE 12.3: Two tabular displays about target genes (top) and evidence (bottom) under the tab Output: target genes.

Section 13 cTCrosstalk

13.1 Interface

Input

• Step 1: a list of user-input SNPs, with 1st column for dbSNP rsIDs and 2nd column for significance info (p-values between 0 and 1). The error message will be displayed if the input is invalid. Example input data are shared genetic variants identified from cross-disease genome-wide association studies in inflammatory disorders; see Nature Genetics 2016.

Mechanism

• Step 2: includes SNPs in Linkage Disequilibrium (LD). By default, input SNPs with a typical threshold (p-value < 5e−8) are considered, and additional SNPs in linkage disequilibrium (R2 < 0.8) can be also included according to the European population.

• Step 3: uses functional genomic datasets, including genomic proximity, quantitative trait locus (QTL) and promoter capture Hi-C (PCHi-C), to identify core genes.

• Step 4: networks core genes with each other and with additional (peripheral) genes based on the knowledge of protein interactions, generating a ranked list of core and peripheral genes. It is achieved using the random walk with restart (RWW) algorithm. By default, the restarting probability of 0.7 is set, empirically optimised for immune-mediated diseases; selecting a value smaller than 0.6 is not recommended as there is a higher chance to expect low performance.

• Step 5: identifies the subnetwork of highly-ranked genes that mediate the crosstalk between molecular pathways. The significance (p-value) of observing the identified crosstalk by chance is estimated by a degree-preserving node permutation test.

• More Controls: fine-tunes parameters involved in steps described above.

Output

• Example Output includes target genes, target pathways, targets at the crosstalk level, and crosstalk-based drug repurposing. A summary of input data and the runtime (computed on the server side) is also returned to the users for the reference.

---

FIGURE 13.1: The interface of the cTCrosstalk, enabling/automating genetics-led and network-based identification and prioritisation of drug targets at the crosstalk level. The Show/Hide Info toggle button contains the help information on how to use the cTCrosstalk, including input, output, mechanism, etc.

13.2 Prioritisation results

• Output: target genes: includes Manhattan plot illustrating priority rating for target genes that are color-coded by chromosomes. Also provided is the downloadable PDF file. It also includes Prioritisation table listing all prioritised genes, each receiving 5-star priority rating (scored 0-5), and Evidence table for core genes showing which SNPs are used to define core genes based on which evidence. Genes are cross-referenced and hyperlinked to GeneCards.

• Output: target pathways: includes a dot plot and a prioritisation table for target pathways. Also provided is the downloadable PDF file.

• Output: targets at the crosstalk level: includes A network visualisation illustrating the crosstalk between pathways, with genes colored by priority rating and labelled in the form of rating®rank, Prioritisation table listing crosstalk genes, each receiving 5-star priority rating (scored 0-5), and Evidence table for pathway crosstalk genes, showing which SNPs are used to crosstalk genes based on which evidence. Genes are cross-referenced and hyperlinked to GeneCards.

• Output: crosstalk-based drug repurposing: includes A heatmap-like illustration showing drug repurposing analysis of approved drugs (licensed medications) based on pathway crosstalk genes, with crosstalk genes on y-axis, disease indications on x-axis, red dots indexed in number and referenced beneath in the table where the information on approved drugs and mechanisms of action is detailed. It also includes An interactive table of crosstalk genes (the column Crosstalk genes), disease indications (the column Disease indications), approved drugs and mechanisms (the column Approved drugs [mechanisms of action]), and drug index (the column Index) shown above within the dot plot.

---

FIGURE 13.2: Prioritisation results for the cTCrosstalk. In addition to a summary of input data and the runtime (computed on the server side) under the tab Input into cTCrosstalk, the prioritisation results page provides the output, including target genes under the tab Output: target genes (the same as shown in the cTGene), target pathways under the tab Output: target pathways, and targets at the crosstalk level under the tab Output: targets at the crosstalk level, and crosstalk-based drug repurposing under the tab Output: crosstalk-based drug repurposing. Under the tab Output: target genes include network visualisation of the crosstalk, with genes/nodes colour-coded by priority rating and labelled in the form of rating®rank, and two tabular displays about prioritisation and evidence for crosstalk genes.

---

FIGURE 13.3: A dot plot for prioritised target pathways, with the top five labelled, available under the tab Output: target pathways. Also available is Prioritisation table for target pathways.

---

FIGURE 13.4: A heatmap-like illustration, with crosstalk genes on the y-axis, disease indications on the x-axis, and red dots indexed in numbers under the tab Output: crosstalk-based drug repurposing. The index numbers are referenced in a table where the information on approved drugs and mechanisms of action is detailed.