---
title: "Target Investigation and Functional Genomics"
author: "Ali Sajid Imami"
date: "`r Sys.Date()`"
output: BiocStyle::html_document
vignette: >
  %\VignetteIndexEntry{Target Investigation and Functional Genomics}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include = FALSE}
knitr::opts_chunk$set(
    collapse = TRUE,
    comment = "#>",
    fig.width = 8L,
    fig.height = 6L,
    warning = FALSE,
    message = FALSE
)
```

# Introduction

The `investigateTarget()` function in **drugfindR** enables comprehensive functional genomics analyses by exploring relationships between genes and drugs in the LINCS database. This vignette covers:

- Understanding what happens when a gene is knocked down or overexpressed
- Finding drugs that mimic or reverse gene perturbation effects
- Discovering genes with similar functional roles
- Elucidating drug mechanisms of action

# Prerequisites

```{r load-libraries}
library(drugfindR)
library(dplyr)
library(ggplot2)
library(tidyr)
library(purrr)
```

# The Target Investigation Workflow

## Key Questions Answered

1. **Gene Function**: What cellular changes occur when a gene is perturbed?
2. **Drug Mimicry**: Which drugs produce effects similar to gene perturbation?
3. **Drug Opposition**: Which drugs counteract gene perturbation effects?
4. **Functional Similarity**: Which genes have similar functional roles?
5. **Mechanism of Action**: What genes does a drug affect?

## Workflow Overview

```
Target (Gene/Drug)
    ↓
Retrieve LINCS Signatures
    ↓
Filter by Magnitude/Direction
    ↓
Query Output Library
    ↓
Generate Consensus
    ↓
Functional Insights
```

# Basic Target Investigation

## Example 1: What does TP53 knockdown do?

TP53 is a tumor suppressor gene. Let's explore what happens when it's knocked down:

```{r tp53-knockdown-basic, eval = FALSE}
# Investigate TP53 knockdown effects
tp53_kd <- investigateTarget(
    target = "TP53",
    inputLib = "KD", # Use knockdown signatures
    outputLib = "KD", # Find similar gene knockdowns
    filterThreshold = 0.5,
    similarityThreshold = 0.3
)

# View results
head(tp53_kd, 15)
```

### Interpreting Results

The output shows:

- **Target**: Other genes whose knockdown produces similar effects
- **Similarity**: Positive values = similar function, negative = opposite function
- **SourceSignature**: TP53 knockdown signature ID(s)
- **TargetSignature**: Matched gene signature IDs

```{r tp53-interpretation, eval = FALSE}
# Find genes with similar function to TP53
similar_genes <- tp53_kd %>%
    filter(Similarity > 0.3) %>%
    arrange(desc(Similarity))

cat("Genes functionally similar to TP53:\n")
head(similar_genes, 10)

# Find genes with opposite function
opposite_genes <- tp53_kd %>%
    filter(Similarity < -0.3) %>%
    arrange(Similarity)

cat("\nGenes with opposite function:\n")
head(opposite_genes, 10)
```

## Example 2: Which drugs mimic TP53 loss?

Cancer cells often lose TP53 function. Which drugs produce similar effects?

```{r tp53-drug-mimics, eval = FALSE}
# Find drugs that mimic TP53 knockdown
tp53_mimics <- investigateTarget(
    target = "TP53",
    inputLib = "KD", # Start with KD signature
    outputLib = "CP", # Find chemical perturbagens
    filterThreshold = 0.5,
    similarityThreshold = 0.3
)

# Drugs with positive similarity mimic TP53 loss
mimicking_drugs <- tp53_mimics %>%
    filter(Similarity > 0.3) %>%
    arrange(desc(Similarity))

head(mimicking_drugs, 10)
```

## Example 3: Which drugs rescue TP53 loss?

More importantly, which drugs might compensate for TP53 loss?

```{r tp53-rescue, eval = FALSE}
# Drugs with negative similarity oppose TP53 loss
rescue_drugs <- tp53_mimics %>%
    filter(Similarity < -0.3) %>%
    arrange(Similarity)

cat("Potential TP53 loss rescue compounds:\n")
head(rescue_drugs, 15)
```

# Gene Overexpression Analysis

## Example 4: Effects of MYC overexpression

MYC is an oncogene frequently overexpressed in cancer:

```{r myc-overexpression, eval = FALSE}
# Investigate MYC overexpression
myc_oe <- investigateTarget(
    target = "MYC",
    inputLib = "OE", # Overexpression signatures
    outputLib = "CP", # Find drugs
    filterThreshold = 0.5
)

# Drugs that oppose MYC overexpression (therapeutic potential)
myc_antagonists <- myc_oe %>%
    filter(Similarity < -0.3) %>%
    arrange(Similarity)

cat("Drugs that oppose MYC overexpression:\n")
head(myc_antagonists, 10)
```

## Comparing KD vs OE

Different perturbation types reveal different biology:

```{r compare-kd-oe, eval = FALSE}
# BRCA1 knockdown
brca1_kd <- investigateTarget(
    target = "BRCA1",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.5
)

# BRCA1 overexpression
brca1_oe <- investigateTarget(
    target = "BRCA1",
    inputLib = "OE",
    outputLib = "CP",
    filterThreshold = 0.5
)

# Compare top drugs
kd_top <- brca1_kd %>%
    filter(Similarity < -0.3) %>%
    pull(Target)
oe_top <- brca1_oe %>%
    filter(Similarity < -0.3) %>%
    pull(Target)

overlap <- intersect(kd_top, oe_top)
cat("Drugs opposing both KD and OE:", length(overlap), "\n")
```

# Drug Mechanism of Action

## Example 5: What does metformin affect?

Reverse the query to understand drug mechanisms:

```{r metformin-moa, eval = FALSE}
# Find genes affected by metformin
metformin_targets <- investigateTarget(
    target = "metformin",
    inputLib = "CP", # Drug signature
    outputLib = "KD", # Find similar gene knockdowns
    filterThreshold = 0.5
)

# Genes whose loss mimics metformin treatment
affected_genes <- metformin_targets %>%
    filter(Similarity > 0.3) %>%
    arrange(desc(Similarity))

cat("Genes potentially targeted by metformin:\n")
head(affected_genes, 15)
```

## Example 6: Compare drug to drug

Find drugs with similar mechanisms:

```{r drug-to-drug, eval = FALSE}
# Find drugs similar to aspirin
aspirin_similar <- investigateTarget(
    target = "aspirin",
    inputLib = "CP",
    outputLib = "CP",
    filterThreshold = 0.5
)

similar_drugs <- aspirin_similar %>%
    filter(Similarity > 0.4) %>%
    arrange(desc(Similarity))

head(similar_drugs, 10)
```

# Paired vs. Unpaired Analysis

## Paired Analysis (Default)

Analyzes up and down-regulated genes separately:

```{r paired-example, eval = FALSE}
# Paired analysis captures directional effects
egfr_paired <- investigateTarget(
    target = "EGFR",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.5,
    paired = TRUE # Default
)

# Results preserve directional information
cat("Directional analysis:\n")
table(egfr_paired$InputSigDirection)
```

### When to use paired:

- Complex genes with many targets
- When directionality matters
- More biologically precise
- Recommended for most analyses

## Unpaired Analysis

Combines all significant genes:

```{r unpaired-example, eval = FALSE}
# Unpaired analysis for simpler interpretation
egfr_unpaired <- investigateTarget(
    target = "EGFR",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.5,
    paired = FALSE
)

# Faster, simpler results
cat("Combined analysis genes:\n")
nrow(egfr_unpaired)
```

### When to use unpaired:

- Exploratory analysis
- Simpler interpretation needed
- Symmetric signatures
- Computational efficiency matters

# Cell Line Considerations

## Filtering Input Cell Lines

Restrict source signatures to specific cell lines:

```{r input-cellline-filtering, eval = FALSE}
# Only use TP53 KD from breast cancer cell lines
tp53_breast <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    inputCellLines = c("MCF7", "MDAMB231", "T47D"),
    filterThreshold = 0.5
)

cat("Results from breast cancer cell lines only\n")
```

## Filtering Output Cell Lines

Restrict target signatures to specific contexts:

```{r output-cellline-filtering, eval = FALSE}
# Find drugs tested in lung cancer cell lines
kras_lung <- investigateTarget(
    target = "KRAS",
    inputLib = "KD",
    outputLib = "CP",
    outputCellLines = c("A549", "H1299", "H460"),
    filterThreshold = 0.5
)

cat("Drugs tested in lung cancer models\n")
```

## Cross-Cell Line Analysis

```{r cross-cellline, eval = FALSE}
# Get all results
all_cells <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.5
)

# Analyze cell line distribution
cellline_summary <- all_cells %>%
    group_by(Target) %>%
    summarize(
        n_celllines = n_distinct(TargetCellLine),
        mean_similarity = mean(Similarity),
        .groups = "drop"
    ) %>%
    filter(n_celllines >= 3) %>%
    arrange(desc(n_celllines))

cat("Drugs tested across multiple cell lines:\n")
head(cellline_summary, 10)
```

# Threshold Optimization

## Conservative Analysis

High confidence, fewer hits:

```{r conservative-thresholds, eval = FALSE}
conservative <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.85, # Very stringent
    similarityThreshold = 0.4 # High similarity required
)

cat("Conservative results:", nrow(conservative), "\n")
```

## Liberal Analysis

Broader coverage, more hypotheses:

```{r liberal-thresholds, eval = FALSE}
liberal <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.3, # Permissive
    similarityThreshold = 0.15 # Lower cutoff
)

cat("Liberal results:", nrow(liberal), "\n")
```

## Comparing Thresholds

```{r threshold-comparison, eval = FALSE}
# Test multiple thresholds
thresholds <- c(0.3, 0.5, 0.7, 0.85)

threshold_results <- map_dfr(thresholds, function(thresh) {
    results <- investigateTarget(
        target = "TP53",
        inputLib = "KD",
        outputLib = "CP",
        filterThreshold = thresh,
        similarityThreshold = 0.2
    )

    tibble(
        threshold = thresh,
        n_results = nrow(results),
        n_strong = sum(abs(results$Similarity) > 0.4)
    )
})

print(threshold_results)
```

# Multiple Target Analysis

## Batch Processing Targets

```{r batch-targets, eval = FALSE}
# Analyze multiple related genes
cell_cycle_genes <- c("TP53", "RB1", "CDKN2A", "MYC", "E2F1")

cell_cycle_results <- map_dfr(cell_cycle_genes, function(gene) {
    tryCatch(
        {
            results <- investigateTarget(
                target = gene,
                inputLib = "KD",
                outputLib = "CP",
                filterThreshold = 0.5
            )

            results %>%
                mutate(SourceGene = gene) %>%
                filter(Similarity < -0.3) # Therapeutic candidates
        },
        error = function(e) {
            message("No results for ", gene)
            tibble()
        }
    )
})

# Find drugs targeting multiple cell cycle genes
multi_target_drugs <- cell_cycle_results %>%
    group_by(Target) %>%
    summarize(
        n_genes_affected = n_distinct(SourceGene),
        mean_similarity = mean(Similarity),
        .groups = "drop"
    ) %>%
    filter(n_genes_affected >= 3) %>%
    arrange(n_genes_affected, mean_similarity)

cat("Drugs affecting multiple cell cycle genes:\n")
head(multi_target_drugs, 10)
```

# Pathway-Level Analysis

## Analyzing Gene Sets

```{r pathway-analysis, eval = FALSE}
# DNA repair pathway genes
dna_repair_genes <- c("BRCA1", "BRCA2", "ATM", "ATR", "CHEK1", "CHEK2")

# Get concordant drugs for each gene
dna_repair_drugs <- map_dfr(dna_repair_genes, function(gene) {
    tryCatch(
        {
            investigateTarget(
                target = gene,
                inputLib = "KD",
                outputLib = "CP",
                filterThreshold = 0.5
            ) %>%
                mutate(PathwayGene = gene)
        },
        error = function(e) tibble()
    )
})

# Drugs consistently affecting the pathway
pathway_drugs <- dna_repair_drugs %>%
    group_by(Target) %>%
    summarize(
        genes_affected = n_distinct(PathwayGene),
        mean_effect = mean(Similarity),
        .groups = "drop"
    ) %>%
    filter(genes_affected >= 4) %>%
    arrange(desc(genes_affected))

cat("Drugs affecting DNA repair pathway:\n")
head(pathway_drugs, 10)
```

# Visualization Strategies

## Similarity Score Distribution

```{r plot-similarity-dist, eval = FALSE}
tp53_results <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.5
)

ggplot(tp53_results, aes(x = Similarity)) +
    geom_histogram(bins = 50, fill = "steelblue", alpha = 0.7) +
    geom_vline(
        xintercept = c(-0.3, 0.3),
        linetype = "dashed",
        color = "red",
        size = 1
    ) +
    labs(
        title = "TP53 Knockdown: Drug Similarity Distribution",
        x = "Similarity Score",
        y = "Number of Drug Signatures",
        subtitle = "Negative = opposes TP53 loss; Positive = mimics TP53 loss"
    ) +
    theme_minimal()
```

## Top Candidates Bar Plot

```{r plot-top-candidates, eval = FALSE}
top_drugs <- tp53_results %>%
    arrange(Similarity) %>%
    head(20)

ggplot(top_drugs, aes(x = reorder(Target, Similarity), y = Similarity)) +
    geom_col(aes(fill = Similarity < 0), show.legend = FALSE) +
    scale_fill_manual(values = c("red3", "steelblue")) +
    coord_flip() +
    labs(
        title = "Top 20 Drugs Related to TP53 Knockdown",
        subtitle = "Blue = opposes TP53 loss (therapeutic potential)",
        x = "Drug/Compound",
        y = "Similarity Score"
    ) +
    theme_minimal() +
    theme(axis.text.y = element_text(size = 10))
```

## Network Visualization Concept

```{r network-prep, eval = FALSE}
# Prepare data for network visualization
# (requires igraph or visNetwork package)

# Get relationships between multiple genes and drugs
network_data <- map_dfr(c("TP53", "BRCA1", "EGFR"), function(gene) {
    investigateTarget(
        target = gene,
        inputLib = "KD",
        outputLib = "CP",
        filterThreshold = 0.5,
        similarityThreshold = 0.3
    ) %>%
        mutate(GeneSource = gene) %>%
        select(GeneSource, Target, Similarity) %>%
        filter(abs(Similarity) > 0.4)
})

# This data can be used with igraph to create gene-drug networks
head(network_data)
```

## Heatmap of Gene-Drug Relationships

```{r heatmap-prep, eval = FALSE}
# Create a matrix of gene-drug similarities
genes_of_interest <- c("TP53", "BRCA1", "MYC", "EGFR", "KRAS")

heatmap_data <- map_dfr(genes_of_interest, function(gene) {
    investigateTarget(
        target = gene,
        inputLib = "KD",
        outputLib = "CP",
        filterThreshold = 0.5
    ) %>%
        mutate(Gene = gene) %>%
        group_by(Gene, Target) %>%
        summarize(Similarity = mean(Similarity), .groups = "drop")
})

# Filter to drugs affecting multiple genes
common_drugs <- heatmap_data %>%
    group_by(Target) %>%
    filter(n() >= 3) %>%
    ungroup()

ggplot(common_drugs, aes(x = Target, y = Gene, fill = Similarity)) +
    geom_tile(color = "white") +
    scale_fill_gradient2(
        low = "blue", mid = "white", high = "red",
        midpoint = 0,
        name = "Similarity"
    ) +
    labs(
        title = "Gene-Drug Relationship Heatmap",
        x = "Drug",
        y = "Gene"
    ) +
    theme_minimal() +
    theme(axis.text.x = element_text(angle = 45, hjust = 1, size = 8))
```

# Advanced Use Cases

## Case 1: Synthetic Lethality Discovery

Find drugs that are specifically toxic when a gene is lost:

```{r synthetic-lethality, eval = FALSE}
# In BRCA1-deficient cells, PARP inhibitors are synthetically lethal
# Find similar relationships

brca1_synthetic <- investigateTarget(
    target = "BRCA1",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.6,
    paired = TRUE
)

# Drugs with strong positive similarity might indicate synergy
potential_synthetic <- brca1_synthetic %>%
    filter(Similarity > 0.4) %>%
    arrange(desc(Similarity))

cat("Potential synthetic lethal partners:\n")
head(potential_synthetic, 10)
```

## Case 2: Rescue vs. Enhancement

```{r rescue-vs-enhance, eval = FALSE}
results <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.5
)

# Classify drugs
drug_classification <- results %>%
    mutate(
        Classification = case_when(
            Similarity < -0.4 ~ "Strong Rescue",
            Similarity < -0.2 ~ "Moderate Rescue",
            Similarity > 0.4 ~ "Strong Enhancement",
            Similarity > 0.2 ~ "Moderate Enhancement",
            TRUE ~ "Neutral"
        )
    )

table(drug_classification$Classification)
```

## Case 3: Temporal Analysis

```{r temporal-analysis, eval = FALSE}
# Analyze time-dependent effects
results <- investigateTarget(
    target = "MYC",
    inputLib = "OE",
    outputLib = "CP",
    filterThreshold = 0.5
)

# Group by time point
temporal_effects <- results %>%
    group_by(TargetTime) %>%
    summarize(
        n_drugs = n_distinct(Target),
        mean_similarity = mean(Similarity),
        .groups = "drop"
    ) %>%
    arrange(TargetTime)

cat("Time-dependent drug effects:\n")
print(temporal_effects)
```

## Case 4: Dose-Response Patterns

```{r dose-response, eval = FALSE}
# Examine concentration-dependent effects
drug_doses <- results %>%
    filter(!is.na(TargetConcentration)) %>%
    group_by(Target, TargetConcentration) %>%
    summarize(
        mean_similarity = mean(Similarity),
        .groups = "drop"
    ) %>%
    group_by(Target) %>%
    filter(n() >= 2) %>% # Multiple concentrations tested
    ungroup()

# Plot dose-response for top drugs
top_drugs_dose <- drug_doses %>%
    group_by(Target) %>%
    slice_head(n = 5) %>%
    ungroup()

ggplot(top_drugs_dose, aes(x = TargetConcentration, y = mean_similarity)) +
    geom_line(aes(color = Target)) +
    geom_point(aes(color = Target), size = 2) +
    labs(
        title = "Dose-Response Patterns",
        x = "Concentration",
        y = "Mean Similarity"
    ) +
    theme_minimal()
```

# Integration with Experimental Data

## Validating Predictions

```{r validation-template, eval = FALSE}
# After getting computational predictions
predictions <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.5
)

# Select top candidates for validation
validation_candidates <- predictions %>%
    filter(Similarity < -0.4, pValue < 0.05) %>%
    arrange(Similarity) %>%
    head(10)

# Export for experimental testing
# write.csv(validation_candidates, "tp53_rescue_candidates.csv")
```

## Comparing with Literature

```{r literature-comparison, eval = FALSE}
# Known TP53-interacting drugs from literature
literature_drugs <- c("nutlin-3", "PRIMA-1", "APR-246")

# Check if they appear in predictions
literature_hits <- predictions %>%
    filter(grepl(paste(literature_drugs, collapse = "|"),
        Target,
        ignore.case = TRUE
    ))

cat("Literature-validated drugs in predictions:\n")
print(literature_hits)
```

# Best Practices

## 1. Start Broad, Then Narrow

```{r practice-funnel, eval = FALSE}
# Step 1: Broad discovery
broad_search <- investigateTarget(
    target = "EGFR",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.5,
    similarityThreshold = 0.2
)

# Step 2: Focus on high-confidence
high_confidence <- broad_search %>%
    filter(abs(Similarity) > 0.4, pValue < 0.01)

# Step 3: Cell line-specific validation
validated <- investigateTarget(
    target = "EGFR",
    inputLib = "KD",
    outputLib = "CP",
    outputCellLines = c("A549", "MCF7"),
    filterThreshold = 0.6,
    similarityThreshold = 0.3
)
```

## 2. Consider Biological Context

- **Cell type**: Use appropriate cell line filters
- **Perturbation type**: KD vs OE have different implications
- **Magnitude**: Balance stringency with coverage
- **Direction**: Paired analysis for complex effects

## 3. Multiple Evidence Lines

```{r multiple-evidence, eval = FALSE}
# Evidence 1: KD → CP
evidence_kd <- investigateTarget(
    target = "BRCA1", inputLib = "KD", outputLib = "CP",
    filterThreshold = 0.5
)

# Evidence 2: OE → CP
evidence_oe <- investigateTarget(
    target = "BRCA1", inputLib = "OE", outputLib = "CP",
    filterThreshold = 0.5
)

# Find drugs appearing in both analyses
drugs_kd <- evidence_kd %>%
    filter(Similarity < -0.3) %>%
    pull(Target)
drugs_oe <- evidence_oe %>%
    filter(Similarity < -0.3) %>%
    pull(Target)

robust_candidates <- intersect(drugs_kd, drugs_oe)
cat("Drugs validated across KD and OE:", length(robust_candidates), "\n")
```

## 4. Document Parameters

Always record your analysis parameters:

```{r documentation, eval = FALSE}
analysis_params <- list(
    target = "TP53",
    input_library = "KD",
    output_library = "CP",
    filter_threshold = 0.5,
    similarity_threshold = 0.3,
    paired = TRUE,
    date = Sys.Date(),
    drugfindR_version = packageVersion("drugfindR")
)

# Save with results
# saveRDS(list(results = tp53_results, params = analysis_params),
#         "tp53_analysis.rds")
```

# Troubleshooting

## No Signatures Found

```{r troubleshoot-none, eval = FALSE}
# If no signatures found for your target:

# 1. Check spelling and case
# Note: drugfindR handles case-insensitivity, but verify gene name

# 2. Try alternate names
# Some genes have aliases (e.g., ERBB2 vs HER2)

# 3. Check metadata
kd_metadata <- drugfindR:::.loadMetadata("KD")
available_genes <- unique(kd_metadata$Source)

# Search for your gene
"TP53" %in% available_genes
grep("TP53", available_genes, value = TRUE, ignore.case = TRUE)
```

## Too Few Results

```{r troubleshoot-few, eval = FALSE}
# If results are sparse:

# 1. Lower thresholds
relaxed <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    filterThreshold = 0.3, # More permissive
    similarityThreshold = 0.1
)

# 2. Try different libraries
# KD → KD instead of KD → CP
alternate <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "KD",
    filterThreshold = 0.5
)

# 3. Remove cell line restrictions
unrestricted <- investigateTarget(
    target = "TP53",
    inputLib = "KD",
    outputLib = "CP",
    inputCellLines = NULL, # All cell lines
    outputCellLines = NULL,
    filterThreshold = 0.5
)
```

# Summary

Key takeaways for target investigation:

1. **Choose right libraries**: KD/OE for genes, CP for drugs
2. **Direction matters**: Positive = similar, negative = opposite
3. **Use paired analysis**: For nuanced biological interpretation
4. **Consider cell context**: Filter by relevant cell lines
5. **Validate predictions**: Multiple evidence lines + experiments
6. **Document thoroughly**: Record all parameters

# Next Steps

- **[Drug Repurposing](drug-repurposing.html)**: Apply to disease signatures
- **[Getting Started](getting-started.html)**: Quick reference guide
- **[Function Reference](https://cogdisreslab.github.io/drugfindR/reference/)**: Complete API

# Session Information

```{r session-info}
sessionInfo()
```

# References

1. Subramanian A, et al. (2017). A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles. *Cell*, 171(6):1437-1452.

2. Lamb J, et al. (2006). The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. *Science*, 313(5795):1929-1935.

3. Corsello SM, et al. (2020). Discovering the anticancer potential of non-oncology drugs by systematic viability profiling. *Nature Cancer*, 1(2):235-248.