FITdb, an Integrated Functional Immunogenomics and Transcriptomics Database

Mapping the Regulatory Logic of Immunity

Modern immunology relies on genetic screens to decipher the complex instructions governing immune cell behavior. Researchers use CRISPR technology—a molecular toolkit to precisely "cut" or modify specific genes—to disable hundreds of genes at once. This helps identify which genes are essential for tasks like killing tumor cells or surviving infections. However, the field faces a massive fragmentation problem. Thousands of these screens are conducted globally, but the resulting datasets are often siloed. They focus on narrow gene panels or specific cell types. Many are also locked away in individual laboratory repositories.

Because these datasets vary in experimental design and technical execution, integration is difficult. A researcher cannot easily ask a single question across multiple studies. One cannot easily determine if a gene important in a mouse T cell experiment is also a critical regulator in a human myeloid cell. This lack of integration prevents the field from extracting broad, generalizable principles of immune regulation. The development of FITdb (the Functional Immunogenomics and Transcriptomics Database) aims to resolve this. It acts as a centralized, harmonized clearinghouse for these disparate genetic worlds.

The Fragmentation of Functional Genomics

The current landscape of functional genomics is characterized by high volume but low interoperability. Researchers typically employ two main modalities to interrogate gene function: pooled screens and single-cell screens. In a pooled screen, a massive library of genetic perturbations is introduced into a population of cells. After a period of selective pressure (such as exposure to a virus), the remaining cells are sequenced. This shows which genetic "cuts" allowed them to survive. This is captured by measuring the Log2 Fold-Change (LFC)—a mathematical ratio describing whether a gene's perturbation led to an enrichment (more cells) or a depletion (fewer cells) of that specific genetic marker [Figure 1A].

While powerful, these pooled screens are often "black boxes" regarding specific cellular states. On the other hand, single-cell CRISPR screens (often called Perturb-seq) allow researchers to see both the genetic perturbation and the entire transcriptional profile (the complete set of RNA molecules being expressed) of an individual cell [Figure 1C]. Despite the wealth of data produced by both methods, the authors note that existing resources are often too specialized. Some focus only on cancer cells or specific metabolic pathways. This leaves a gap for a holistic, cross-species, and cross-modality resource.

Harmonizing Diverse Genetic Datasets

To bridge this gap, the authors built a relational architecture designed for comparison. The construction of FITdb follows a rigorous three-stage logic:

1. Standardized Re-analysis: Rather than relying on heterogeneous processing pipelines, the researchers uniformly re-analyzed the data. Whenever possible, they started from the original FASTQ files (raw sequencing reads). They also used standardized count tables generated by MAGeCK, a software tool for identifying essential genes from CRISPR data.
2. Multi-Modal Integration: The database integrates 43 independent screens. This includes 32 pooled screens and 11 single-cell screens [Figure 1D]. This allows the platform to handle both the "bulk" population-level shifts of pooled assays and the high-resolution "cellular snapshots" provided by Perturb-seq.
3. Relational Mapping: The processed data is housed in a MariaDB database. This allows for complex queries that link specific single-guide RNAs (sgRNAs, the short RNA sequences that guide the CRISPR machinery to a target gene) directly to biological outcomes like gene expression or cell fitness.

The architecture moves from the microscopic to the systemic. Through modules like the "Gene Report," a user can drill down from a single gene to its specific sgRNA statistics. They can then move upward through "Pathway Analysis" to see how those genes cluster into broader biological networks, such as inflammatory responses [Figure 2C-G].

Scale and Statistical Rigor

The scale of FITdb is its most defining characteristic. The authors report that the database integrates 20,696 mouse genes and 22,293 human genes. This coverage spans 199 different immune cell types and experimental conditions [Figure 1D-F]. This scale transforms the database from a simple lookup table into a comparative engine.

One significant technical contribution is the "Compare MyGeneSet" tool. Instead of simple list-matching, the authors implement a one-sided hypergeometric test (looking at the upper tail) to determine statistical significance. This determines if a user's custom list of genes is over-represented within the curated functional gene sets of FITdb [Figure 3B]. A researcher can take a list of genes from a novel experiment and immediately ask if they are significantly similar to genes known to regulate T cell exhaustion in existing studies.

The utility of this integration is demonstrated in a use case involving the adaptors LAT and LAT2. By using the "Cross-Screen Visualizer," the authors showed that while LAT promotes T cell activation, the closely related LAT2 actually acts as a restraint on productive signaling [Figure 4A]. This distinction was only possible by viewing the data through a lens that spans multiple modalities and experimental contexts.

Limitations of the Current Integration

Despite its breadth, FITdb is not a complete map of the immune system. A primary limitation is that transcriptomic data (the detailed measurement of gene expression) is currently only available for a subset of the integrated bulk screens. For the remaining pooled screens, users can see the "fitness" outcome but cannot see the underlying changes in the cell's molecular state.

Furthermore, while the database is effective at identifying what happens when a gene is perturbed, it remains a descriptive resource. It provides evidence for a regulatory relationship. However, it does not solve the fundamental biochemical "how"—the physical mechanism of protein interaction. Finally, the database's utility is tied to the availability of public data. As new screens are published, the database must be curated and updated to maintain its relevance.

The Verdict: A Necessary Infrastructure

FITdb represents a step toward making functional immunology a truly integrative science. It moves away from isolated, single-study snapshots. Instead, it offers a harmonized, multi-modal landscape. For practitioners, the tool is a high-leverage asset for validating new CRISPR hits. For theorists, it provides a way to discover the conserved logic of immune regulation.

The authors plan to expand the database to host more than 80 screens by the end of 2027. They also intend to develop predictive imputation methods for "unseen" genes. Consequently, FITdb is transitioning from a static repository to a dynamic model of the immune system. The resource is freely available at https://fitdb.lji.org.

Figures from the paper