Functional Annotation in Bioinformatics: KEGG, GO, EggNOG, InterPro, Pfam, CAZy, CARD, and Beyond

After assembling a genome, building a metagenome, predicting genes, or receiving a list of proteins, the next question is usually simple:

What do these genes actually do?

The answer is not simple.

A gene may encode an enzyme, contain a conserved domain, belong to an orthologous group, participate in a pathway, contribute to virulence, confer antimicrobial resistance, or perform a function that is still unknown.

This is why functional annotation is not a single-database problem.

KEGG is useful, but it is only one layer. UniProt is useful, but it is not enough. For microbial genomes, metagenomes, MAGs, viral genomes, pangenomes, and non-model organisms, a good annotation workflow usually combines several resources.

This guide explains the major functional annotation databases, how they relate to each other, and how to use them in practical workflows.

What Functional Annotation Means
The Functional Annotation Stack
Database Overview
UniProt
KEGG and KEGG Orthology
KEGG BRITE Hierarchies
Gene Ontology
eggNOG and COG Categories
Pfam and Protein Domains
InterPro and InterProScan
CAZy and dbCAN
CARD and Antimicrobial Resistance
VFDB and Virulence Factors
MEROPS and Proteases
TCDB and Transporters
MetaCyc and Reactome
Functional Annotation Workflows
Metagenomes and MAGs
Viral Genomes
Non-model Organisms
Practical KEGG BRITE Mapping Workflow
Choosing the Right Tool
Common Pitfalls
Recommended Output Tables
Example Project Structure
Practical Interpretation Strategy

What Functional Annotation Means

Functional annotation means assigning biological meaning to sequences.

A sequence may be annotated at several levels:

Sequence
↓
Predicted gene
↓
Protein sequence
↓
Domain
↓
Protein family
↓
Ortholog group
↓
Enzyme reaction
↓
Pathway
↓
Biological process

Different databases answer different questions.

For example, given one protein, you may ask:

What is its likely name?
Does it contain a known domain?
Is it part of a protein family?
Does it belong to an ortholog group?
Does it catalyze a reaction?
Is it part of a pathway?
Is it related to virulence?
Is it associated with antimicrobial resistance?
Is it a transporter?
Is it a carbohydrate-active enzyme?
Is it conserved across species?

No single database answers all of these well.

The Functional Annotation Stack

Many beginners imagine annotation like this:

Gene
↓
Function

In real projects, it is closer to this:

DNA sequence
↓
Gene prediction
↓
Protein sequence
↓
Similarity search
↓
Domain search
↓
Orthology assignment
↓
Functional classification
↓
Pathway reconstruction
↓
Biological interpretation

The distinction matters because different tools operate at different levels.

For example:

BLAST gives similarity.
Pfam gives domains.
InterPro integrates multiple signature databases.
EggNOG gives orthology and broad functional categories.
KEGG gives KO IDs and pathways.
GO gives biological vocabulary.
CAZy describes carbohydrate-active enzymes.
CARD describes antimicrobial resistance genes.
VFDB describes virulence factors.

A useful annotation is usually the result of combining these layers.

Database Overview

Resource	Main Use	Typical Output
UniProt	Protein knowledgebase	Protein names, functions, GO, cross-references
KEGG	Pathways and orthology	KO IDs, pathways, modules
KEGG BRITE	Hierarchical classification	Functional hierarchy
GO	Controlled functional vocabulary	GO terms
EggNOG	Orthologous groups	Orthologs, COG categories, GO, KO
COG	Broad evolutionary categories	COG letters
Pfam	Protein domains	Domain families
InterPro	Integrated protein signatures	Domains, families, GO terms
TIGRFAM	Curated protein families	Microbial protein families
CAZy	Carbohydrate-active enzymes	GH, GT, PL, CE, AA, CBM families
dbCAN	Automated CAZy annotation	CAZyme predictions
CARD	Antimicrobial resistance	AMR gene families
ResFinder	Antimicrobial resistance	Resistance genes
VFDB	Virulence factors	Virulence-associated genes
MEROPS	Proteases	Peptidase families
TCDB	Transporters	Transporter classification
MetaCyc	Metabolic pathways	Reactions and pathways
Reactome	Curated pathways	Human and model-organism pathways

UniProt

UniProt is a protein knowledgebase.

It is often the first place people look when they want to know what a protein does.

UniProt contains:

Protein names
Functional descriptions
Domain information
Enzyme classification
GO terms
Cross-references to KEGG, Pfam, InterPro, Reactome, and other resources
Literature-supported annotations

UniProt has two major sections:

Section	Meaning
Swiss-Prot	Manually curated
TrEMBL	Automatically annotated

Swiss-Prot is usually more reliable but much smaller. TrEMBL is much larger but noisier.

When UniProt Is Useful

UniProt is useful when you have proteins that are close to known, curated proteins.

It is especially helpful for:

Model organisms
Well-studied enzymes
Human proteins
Bacterial proteins with strong homologs
Manual review of important genes

When UniProt Is Not Enough

UniProt is less sufficient for:

Environmental metagenomes
Novel viral proteins
Highly divergent proteins
MAGs from poorly characterized clades
Non-model organisms with sparse annotation

In those cases, domain-based and orthology-based annotation often becomes more useful.

KEGG and KEGG Orthology

KEGG is widely used for pathway and metabolism analysis.

The most important KEGG concept for functional annotation is the KO ID.

KO means KEGG Orthology.

A KO groups genes that perform equivalent biological functions, even if they come from different organisms.

Example:

Human hexokinase
Yeast hexokinase
Bacterial glucokinase
↓
K00844

The organisms are different, but the functional role is comparable.

This is why KEGG is powerful in metagenomics and comparative genomics.

Common KEGG Identifiers

Identifier	Meaning
K number	KEGG Orthology
PATH	Pathway
MODULE	Functional module
EC	Enzyme Commission number
BR	BRITE hierarchy

Example:

K00844
HK; hexokinase [EC:2.7.1.1]
Glycolysis / Gluconeogenesis [PATH:ko00010]

When KEGG Is Useful

KEGG is useful for:

Metabolism
Pathway reconstruction
Microbial genome interpretation
Metagenomic functional profiling
Comparing pathway potential between samples

When KEGG Is Limited

KEGG may be less complete for:

Novel proteins
Poorly characterized organisms
Some viral proteins
Regulatory functions outside curated pathways
Specialized databases like AMR or virulence

KEGG BRITE Hierarchies

KEGG BRITE provides hierarchical classification.

Instead of listing thousands of KO IDs, BRITE allows you to summarize them into higher-level categories.

Example:

Metabolism
└── Carbohydrate metabolism
    └── Glycolysis / Gluconeogenesis
        └── K00844
            HK; hexokinase

This is useful because raw KO tables are difficult to interpret.

A table like this:

Gene	KO
gene_001	K00844
gene_002	K01810
gene_003	K01689

can be summarized into:

Level 1	Level 2	Count
Metabolism	Carbohydrate metabolism	3

That is easier to explain in a paper, report, or figure.

Gene Ontology

Gene Ontology, or GO, provides a controlled vocabulary for gene functions.

GO has three branches.

Biological Process

What larger biological process is involved?

Examples:

glycolysis
DNA repair
cell division
immune response

Molecular Function

What does the molecule do?

Examples:

ATP binding
kinase activity
DNA binding
oxidoreductase activity

Cellular Component

Where does it act?

Examples:

ribosome
cytoplasm
mitochondrion
plasma membrane

Example: Hexokinase

A hexokinase-like protein may have annotations such as:

GO:0004396    hexokinase activity
GO:0006096    glycolytic process
GO:0005737    cytoplasm

Each GO term describes a different aspect of function.

Why GO Is Useful

GO is useful because it is:

Structured
Searchable
Compatible with enrichment analysis
Used across many organisms
Supported by many annotation tools

GO is often used downstream for:

Enrichment analysis
Functional summaries
Gene set interpretation
Comparative analyses

EggNOG and COG Categories

EggNOG is one of the most useful resources for microbial and metagenomic annotation.

EggNOG Mapper can annotate proteins with:

Orthologous groups
Functional descriptions
GO terms
KEGG KO IDs
EC numbers
COG categories
Preferred names

Example command:

emapper.py \
  -i proteins.faa \
  --itype proteins \
  --output annotation \
  --cpu 8

Typical output columns include:

query
seed_ortholog
evalue
score
eggNOG_OGs
max_annot_lvl
COG_category
Description
Preferred_name
GOs
EC
KEGG_ko
KEGG_Pathway

COG Categories

COG categories are broad functional classes.

Code	Category
J	Translation, ribosomal structure and biogenesis
A	RNA processing and modification
K	Transcription
L	Replication, recombination and repair
B	Chromatin structure and dynamics
D	Cell cycle control
V	Defense mechanisms
T	Signal transduction
M	Cell wall, membrane, envelope biogenesis
N	Cell motility
U	Intracellular trafficking and secretion
O	Post-translational modification
C	Energy production and conversion
G	Carbohydrate transport and metabolism
E	Amino acid transport and metabolism
F	Nucleotide transport and metabolism
H	Coenzyme transport and metabolism
I	Lipid transport and metabolism
P	Inorganic ion transport and metabolism
Q	Secondary metabolite biosynthesis
R	General function prediction only
S	Function unknown

These categories are common in microbial genome papers.

A common figure is:

COG category abundance per genome

or:

COG category abundance per metagenomic bin

Pfam and Protein Domains

Pfam classifies protein domains.

A domain is a reusable functional or structural unit inside a protein.

For example:

PF00069
Protein kinase domain

A protein can contain multiple domains.

Example:

Signal peptide
↓
Transmembrane domain
↓
Catalytic domain
↓
Binding domain

Domain-based annotation is useful when full-length similarity is weak.

This is common in:

Non-model organisms
Environmental metagenomes
Viral genomes
Fragmented assemblies
Distant homologs

Pfam helps answer:

What does this part of the protein look like?

not always:

What is the full biological role of this protein?

InterPro and InterProScan

InterPro integrates multiple protein signature databases.

It includes resources such as:

Pfam
SMART
PROSITE
TIGRFAM
Gene3D
SUPERFAMILY
CDD
PIRSF

InterProScan is the tool used to run these searches.

Example:

interproscan.sh \
  -i proteins.faa \
  -f tsv \
  -dp \
  -cpu 8

Typical outputs include:

Protein ID
Database
Signature ID
Description
Start
End
Score
InterPro ID
GO terms
Pathway annotations

When InterProScan Is Useful

InterProScan is useful when you want comprehensive domain and family annotation.

It is especially helpful for:

Non-model organism annotation
Genome annotation
Protein family discovery
GO term assignment
Manual investigation of unknown proteins

Trade-off

InterProScan can be computationally heavy.

For small to moderate protein sets, it is excellent.

For huge metagenomic catalogs, EggNOG Mapper or targeted HMM searches may be more practical.

CAZy and dbCAN

CAZy classifies carbohydrate-active enzymes.

These are enzymes that build, break, or modify carbohydrates.

Major CAZy classes include:

Class	Meaning
GH	Glycoside hydrolases
GT	Glycosyltransferases
PL	Polysaccharide lyases
CE	Carbohydrate esterases
AA	Auxiliary activities
CBM	Carbohydrate-binding modules

Examples:

Family	Common Function
GH13	Alpha-amylases
GH18	Chitinases
GH5	Cellulases
GT2	Cellulose synthases
AA2	Lignin-modifying enzymes

CAZy is important in:

Gut microbiomes
Rumen microbiomes
Soil microbiology
Marine microbiology
Plant pathogens
Aquaculture systems
Biomass degradation

dbCAN

dbCAN is commonly used for automated CAZyme annotation.

Example:

run_dbcan proteins.faa protein \
  --out_dir dbcan_out

dbCAN is often better than relying on general-purpose annotation for carbohydrate metabolism.

CARD and Antimicrobial Resistance

For antimicrobial resistance, use specialized databases.

KEGG or GO may tell you that something is a beta-lactamase-like enzyme, but AMR interpretation usually needs more detail.

CARD, the Comprehensive Antibiotic Resistance Database, is commonly used.

Common AMR genes include:

blaTEM
blaCTX-M
tetA
tetM
vanA
mecA

Typical tool:

rgi main \
  --input_sequence proteins.faa \
  --output_file card_annotation \
  --input_type protein

CARD helps identify:

Resistance genes
Resistance mechanisms
Drug classes
Homology models
SNP-based resistance markers in some cases

Important Warning

AMR annotation should be interpreted carefully.

A match to a resistance-related protein does not always mean the organism is clinically resistant.

Important factors include:

Identity
Coverage
Gene completeness
Expression
Genomic context
Known resistance mechanism
Phenotypic validation

VFDB and Virulence Factors

Virulence factors are genes that contribute to pathogenicity.

VFDB is commonly used for virulence annotation.

Examples:

Type III secretion system
Type VI secretion system
hemolysin
adhesin
flagellar proteins
toxins
capsule biosynthesis genes

VFDB is useful in:

Clinical microbiology
Foodborne pathogen analysis
Aquaculture pathogen studies
Comparative pathogenomics

Important Warning

Virulence annotation is context-dependent.

A gene similar to a virulence factor does not automatically mean the organism is pathogenic.

Interpretation depends on:

Host
Organism
Gene completeness
Expression
Genomic island context
Known biology

MEROPS and Proteases

MEROPS classifies peptidases and protease inhibitors.

Proteases are important in:

Host-pathogen interactions
Tissue invasion
Digestion
Immune evasion
Protein maturation

MEROPS is useful when protease function matters more than broad pathway annotation.

TCDB and Transporters

The Transporter Classification Database classifies membrane transport proteins.

Transporters are often important but overlooked.

Examples:

ABC transporters
MFS transporters
ion channels
metal transporters
drug efflux pumps
sugar transporters

Transporter annotation matters in:

Nutrient acquisition
Environmental adaptation
Antimicrobial resistance
Host colonization
Metal tolerance

MetaCyc and Reactome

MetaCyc and Reactome are pathway databases.

MetaCyc

MetaCyc is useful for metabolic pathway reconstruction.

It is especially useful in microbial and plant metabolism.

Reactome

Reactome is highly curated and strongest for human and model organism pathways.

It is useful for:

Human biology
Signaling pathways
Disease biology
Systems biology

For non-human metagenomes, KEGG, MetaCyc, and EggNOG are usually more common.

Functional Annotation Workflows

A general genome annotation workflow:

Assembly
↓
Gene prediction
↓
Protein FASTA
↓
EggNOG Mapper
↓
InterProScan
↓
KEGG / KofamScan
↓
Specialized databases
↓
Summary tables

A minimal practical workflow:

proteins.faa
↓
EggNOG Mapper
↓
Functional summary

A more complete workflow:

proteins.faa
├── EggNOG Mapper
├── InterProScan
├── KofamScan
├── dbCAN
├── CARD
└── VFDB

Then merge the results into one annotation table.

Metagenomes and MAGs

Functional annotation is especially important in metagenomics.

A typical metagenomic workflow:

Raw reads
↓
Quality control
↓
Assembly
↓
Gene prediction
↓
Protein FASTA
↓
Functional annotation
↓
Gene abundance
↓
Pathway abundance

For MAGs:

Assembly
↓
Binning
↓
MAG quality check
↓
Gene prediction
↓
Functional annotation
↓
Metabolic reconstruction

Common biological questions:

Can this organism degrade cellulose?
Can it metabolize sulfur?
Can it fix nitrogen?
Can it use methane?
Can it produce short-chain fatty acids?
Does it carry AMR genes?
Does it encode secretion systems?
Does it have complete biosynthetic pathways?

Common MAG Annotation Tools

Tool	Use
Prodigal	Gene prediction
Prokka	Prokaryotic annotation
Bakta	Modern bacterial genome annotation
EggNOG Mapper	Orthology and function
DRAM	Metabolic annotation of genomes and MAGs
KofamScan	KEGG KO assignment
dbCAN	CAZyme annotation

Viral Genomes

Viral annotation is difficult.

Many viral proteins have no known function.

Common issues:

High sequence diversity
Small genomes
Many hypothetical proteins
Few experimentally validated proteins
Weak similarity to known proteins

A practical viral annotation workflow:

Viral genome
↓
ORF prediction
↓
Protein FASTA
↓
BLAST / DIAMOND
↓
HMM search
↓
InterProScan
↓
VOG / viral protein databases
↓
Manual curation

For viral genomes, it is normal for many proteins to remain:

hypothetical protein

Do not over-annotate weak hits.

Non-model Organisms

Non-model organisms often have sparse annotations.

A direct BLAST hit may produce vague labels such as:

uncharacterized protein
hypothetical protein
predicted protein

In these cases, domain and orthology tools are especially useful.

Better strategy:

Protein sequence
↓
InterPro / Pfam
↓
EggNOG
↓
KEGG
↓
GO
↓
Manual review

For non-model organisms, avoid assuming that a weak similarity hit gives a precise function.

Practical KEGG BRITE Mapping Workflow

The original version of this guide focused on using UniProt and KEGG BRITE to classify proteins hierarchically.

That workflow is still useful when you want to map UniProt IDs to KEGG KO IDs and then classify them into BRITE categories.

Download UniProt Swiss-Prot

wget https://ftp.uniprot.org/pub/databases/uniprot/current_release/knowledgebase/complete/uniprot_sprot.fasta.gz

gunzip uniprot_sprot.fasta.gz

Extract UniProt accessions:

awk '/^>/ {
    match($0, /\|([^|]+)\|/, a)
    print a[1]
}' uniprot_sprot.fasta > uniprot_ids.txt

Map UniProt IDs to KEGG Genes and KO IDs

KEGG API access should be used politely. Avoid excessive parallel requests.

cat uniprot_ids.txt \
| xargs -P 3 -I {} bash -c '
    uid="{}"
    conv=$(curl -s -L "https://rest.kegg.jp/conv/genes/uniprot:${uid}")
    gene=$(printf "%s\n" "$conv" | awk -F "\t" "NR==1 {print \$2}")
    if [ -n "$gene" ]; then
        curl -s -L "https://rest.kegg.jp/link/ko/${gene}" \
        | awk -v uid="$uid" -v gene="$gene" "BEGIN {FS=OFS=\"\t\"} {print uid, gene, \$2}"
    fi
' > uniprot_gene_ko.tsv

Expected columns:

UniProt_ID
KEGG_Gene_ID
KO_ID

Example:

P19367    hsa:3098    ko:K00844

Download KEGG BRITE KO Hierarchy

curl -L https://rest.kegg.jp/get/br:ko00001/json \
  -o ko00001.json

Convert KEGG BRITE JSON to a TSV Table

For production use, Python is more maintainable than a long shell one-liner.

Save this as parse_kegg_brite.py:

#!/usr/bin/env python3

import json
import re
import sys

def walk(node, path):
    name = node.get("name", "")
    children = node.get("children", [])

    new_path = path + [name]

    if not children:
        if name.startswith("K"):
            ko = name.split()[0]
            desc = " ".join(name.split()[1:])
            levels = path[:3]
            while len(levels) < 3:
                levels.append("")
            print("\t".join([ko] + levels + [desc]))
        return

    for child in children:
        walk(child, new_path)

with open(sys.argv[1]) as handle:
    data = json.load(handle)

print("KO_ID\tLevel_1\tLevel_2\tLevel_3\tDescription")

for child in data.get("children", []):
    walk(child, [])

Run:

python parse_kegg_brite.py ko00001.json > brite.tsv

Example output:

KO_ID	Level_1	Level_2	Level_3	Description
K00844	09100 Metabolism	09101 Carbohydrate metabolism	00010 Glycolysis / Gluconeogenesis	HK; hexokinase [EC:2.7.1.1]
K01810	09100 Metabolism	09101 Carbohydrate metabolism	00010 Glycolysis / Gluconeogenesis	GPI, pgi; glucose-6-phosphate isomerase [EC:5.3.1.9]

Clean BRITE Labels

awk 'BEGIN {FS=OFS="\t"}
NR==1 {print; next}
{
    for (i=2; i<=4; i++) {
        sub(/^[0-9]+ /, "", $i)
    }
    print
}' brite.tsv > brite.clean.tsv

Merge UniProt KO Table with BRITE Table

awk 'BEGIN {FS=OFS="\t"}
FNR==NR {
    brite[$1]=$2 OFS $3 OFS $4 OFS $5
    next
}
{
    ko=$3
    sub(/^ko:/, "", ko)
    if (ko in brite) {
        print $1, $2, ko, brite[ko]
    }
}' brite.clean.tsv uniprot_gene_ko.tsv \
> uniprot_brite.tsv

Expected output:

UniProt_ID	KEGG_Gene_ID	KO_ID	Level_1	Level_2	Level_3	Description
P19367	hsa:3098	K00844	Metabolism	Carbohydrate metabolism	Glycolysis / Gluconeogenesis	HK; hexokinase [EC:2.7.1.1]

When This Workflow Is Useful

This workflow is useful when:

You already have UniProt IDs
You want KO IDs
You want hierarchical KEGG BRITE summaries
You want a reusable annotation table
You are preparing pathway-level summaries

When This Workflow Is Not Enough

It is not enough when:

Your proteins do not map well to UniProt
You are working with novel metagenomic proteins
You need domain-level annotation
You need AMR or virulence annotation
You need CAZyme annotation
You need genome-scale metabolic reconstruction

In those cases, combine KEGG with EggNOG, InterPro, dbCAN, CARD, VFDB, or DRAM.

Choosing the Right Tool

Goal	Recommended Resource
General protein annotation	UniProt, EggNOG
Orthology	EggNOG, KEGG KO
Pathways	KEGG, MetaCyc, Reactome
GO terms	UniProt, InterPro, EggNOG
Protein domains	Pfam, InterPro
Microbial genomes	EggNOG, Bakta, DRAM
MAGs	DRAM, EggNOG, KofamScan
CAZymes	dbCAN, CAZy
AMR genes	CARD, ResFinder
Virulence	VFDB
Transporters	TCDB
Proteases	MEROPS
Viral proteins	HMMs, InterPro, viral databases

Common Pitfalls

Treating Similarity as Proof of Function

A BLAST hit does not prove function.

Weak similarity may indicate:

Shared domain
Distant homology
Partial match
Incorrect database annotation
Conserved but unknown protein family

Always check identity, coverage, and biological plausibility.

Over-interpreting Hypothetical Proteins

Many proteins are annotated as:

hypothetical protein
uncharacterized protein
DUF-containing protein

This is not a failure. It is a realistic result, especially for viruses and environmental data.

Ignoring Coverage

A 95% identity hit across 20% of a protein is not the same as a 95% identity hit across 95% of the protein.

Functional transfer should consider:

Percent identity
Query coverage
Subject coverage
Domain boundaries
Alignment quality

Mixing Annotation Levels

These are not equivalent:

PF00069
K00844
GO:0004672
EC:2.7.11.1

They describe different things:

Identifier	Meaning
Pfam	Domain
KEGG KO	Orthologous function
GO	Controlled function term
EC	Enzyme reaction
COG	Broad category

Do not treat them as interchangeable.

Assuming Human Pathways Apply Everywhere

Reactome and many curated pathway resources are strongest for human biology.

For bacteria, archaea, metagenomes, and MAGs, KEGG, MetaCyc, EggNOG, and DRAM are often more appropriate.

Ignoring Specialized Databases

General annotation may miss important biology.

For example:

CAZymes require CAZy/dbCAN.
AMR requires CARD/ResFinder.
Virulence requires VFDB.
Transporters require TCDB.
Proteases require MEROPS.

Recommended Output Tables

A useful functional annotation table should have one row per gene or protein.

Suggested columns:

gene_id
contig_id
start
end
strand
protein_id
product
preferred_name
description
ko_id
kegg_pathway
kegg_brite_level_1
kegg_brite_level_2
kegg_brite_level_3
go_terms
ec_number
cog_category
eggnog_og
pfam
interpro
cazy_family
card_hit
vfdb_hit
tcdb_hit
best_hit
best_hit_identity
best_hit_coverage
annotation_source

For many projects, this table becomes the central annotation file.

Example Project Structure

annotation_project/
├── input/
│   ├── assembly.fna
│   └── proteins.faa
├── eggnog/
│   └── annotation.emapper.annotations
├── interpro/
│   └── interproscan.tsv
├── kegg/
│   ├── kofamscan.tsv
│   └── brite.clean.tsv
├── dbcan/
│   └── overview.txt
├── card/
│   └── card_annotation.txt
├── vfdb/
│   └── vfdb_hits.tsv
├── merged/
│   └── functional_annotation.tsv
└── README.md

A clean directory structure prevents annotation projects from turning into a pile of unrelated TSV files.

Practical Interpretation Strategy

When reviewing an important gene, do not rely on only one line of annotation.

Use a layered approach:

What is the best sequence similarity hit?
What domains are present?
Is there a KEGG KO?
Are there GO terms?
Is there an EC number?
Is it part of a pathway?
Is it found in related organisms?
Is the genomic neighborhood informative?
Is the annotation supported by multiple databases?
Is manual curation needed?

The more layers agree, the more confident the annotation becomes.

Table of Contents

What Functional Annotation Means

The Functional Annotation Stack

Database Overview

UniProt

When UniProt Is Useful

When UniProt Is Not Enough

KEGG and KEGG Orthology

Common KEGG Identifiers

When KEGG Is Useful

When KEGG Is Limited

KEGG BRITE Hierarchies

Gene Ontology

Biological Process

Molecular Function

Cellular Component

Example: Hexokinase

Why GO Is Useful

EggNOG and COG Categories

COG Categories

Pfam and Protein Domains

InterPro and InterProScan

When InterProScan Is Useful

Trade-off

CAZy and dbCAN

dbCAN

CARD and Antimicrobial Resistance

Important Warning

VFDB and Virulence Factors

Important Warning

MEROPS and Proteases

TCDB and Transporters

MetaCyc and Reactome

MetaCyc

Reactome

Functional Annotation Workflows

Metagenomes and MAGs

Common MAG Annotation Tools

Viral Genomes

Non-model Organisms

Practical KEGG BRITE Mapping Workflow

Download UniProt Swiss-Prot

Map UniProt IDs to KEGG Genes and KO IDs

Download KEGG BRITE KO Hierarchy

Convert KEGG BRITE JSON to a TSV Table

Clean BRITE Labels

Merge UniProt KO Table with BRITE Table

When This Workflow Is Useful

When This Workflow Is Not Enough

Choosing the Right Tool

Common Pitfalls

Treating Similarity as Proof of Function

Over-interpreting Hypothetical Proteins

Ignoring Coverage

Mixing Annotation Levels

Assuming Human Pathways Apply Everywhere

Ignoring Specialized Databases

Recommended Output Tables

Example Project Structure

Practical Interpretation Strategy