🧬 Multi-Omics Architecture DNA • RNA • Protein • Metabolite Systems Biology Framework

The Multi-Omics Landscape: Genomics, Transcriptomics, Proteomics & Metabolomics

Biological systems operate through dynamic, multi-layered regulatory networks. Explore comprehensive guides, technologies, computational toolchains, and data architectures across every major omics discipline—from the genomic blueprint to the metabolic phenotype.

🧬 Genomics 📊 Transcriptomics 🔬 Proteomics 🧪 Metabolomics 🧬 Epigenomics 🦠 Metagenomics 🌱 Plant & Polyploid 🌐 Multi-Omics Integration

🔬 MULTI-OMICS PRODUCTION MATRIX · v3.2

Functional & Regulatory Omics Workbench

Select an omic layer to inspect its end-to-end tool stack, statistical normalization model, and toy dataset.

Raw Input & Conversion

Thermo .raw / Bruker .d → .mzML

ThermoRawFileParser / msconvert

Primary Quantification Engine

DIA-NN / FragPipe (MSFragger)

In-silico spectral lib + MaxLFQ MBR

Normalization & Statistics

VSN + MinProb/QRILC + Limma

Bioconductor DEP / MSstats / KSEA

Practice Dataset & Specs

proteomics_mini_diann.tsv (4.2 MB)

⚡ 16–32 Cores · 64 GB RAM

🔄 The Central Dogma & The Hierarchy of Life

Each omics layer captures a unique phase in the flow of biological information. Combining these orthogonal dimensions provides an unprecedented systems-level view of health and disease:

Figure 1.1: The Multi-Omics Continuum of Life — Orthogonal information flow from genomic sequence to metabolic phenotype, orchestrated by transcriptional regulation, ribosomal translation, and biochemical pathway flux.

Layer 1: The Blueprint

Genomics (DNA)

What can happen: Static genetic code, germline mutations, somatic variants, and structural rearrangements.

Layer 2: The Message

Transcriptomics (RNA)

What appears to be happening: Dynamic gene expression levels, alternative splicing, non-coding RNAs, and single-cell states.

Layer 3: The Machinery

Proteomics (Proteins)

What is doing the work: Protein abundance, 3D structures, post-translational modifications (PTMs), and enzymatic complexes.

Layer 4: The Phenotype

Metabolomics (Small Mol.)

What has actually happened: Cellular metabolites, biochemical pathway flux, lipids, sugars, and environmental readouts.

🧬 1. Genomics: Deciphering the DNA Blueprint

DNA • Variants • Assembly

Genomics investigates the complete genetic complement of an organism. Modern computational genomics revolves around high-throughput sequencing (NGS), aligning billions of short or long reads to human reference genomes (GRCh38, T2T-CHM13), and discovering genetic variations that govern disease, development, and evolution.

Figure 1.2: High-Throughput Sequencing Chemistries & Variant Detection Spectrum — Short-read Illumina SBS cluster optics, long-read single-molecule real-time architectures, and resolution of variant classes from SNVs to multi-megabase structural translocations.

Sequencing Platforms

Illumina NovaSeq X (2×150 bp short reads, Q30+ accuracy), Pacific Biosciences (PacBio) HiFi long reads (15-20 kb, >99.9% accuracy), and Oxford Nanopore (ONT) ultra-long reads for resolving complex repeats and telomeres.

Compare NGS Technologies →

Variant Discovery (SNVs, Indels, SVs)

Standardized alignment with BWA-MEM2 or Minimap2, coordinate sorting and duplicate marking with samtools, and variant calling using GATK HaplotypeCaller or Google DeepVariant.

Practical Genomics Track →

De Novo Genome Assembly

Reconstructing complete chromosomal contigs using overlap-layout-consensus (OLC) or de Bruijn graphs (hifiasm, Flye, SPAdes) combined with optical mapping and Hi-C scaffolding.

Genome Assembly Guide →

📊 2. Transcriptomics: Profiling the Expressed Genome

RNA-Seq • Single-Cell • Spatial

Transcriptomics captures the quantity and identity of all RNA transcripts (mRNA, lncRNA, miRNA) active in a biological specimen at a given snapshot in time. It measures dynamic responses to environmental stimuli, drugs, and genetic perturbations.

Figure 1.3: Transcriptomics Quantification Modes — Bulk RNA-Seq exon splicing coverage models (Sashimi plot) juxtaposed with single-cell droplet microfluidic barcoding and 2D UMAP non-linear manifold clustering.

Bulk RNA-Seq & Differential Expression

Splice-aware alignment with STAR or pseudo-alignment with Salmon/Kallisto. Negative binomial dispersion modeling and gene shrinkage with DESeq2 and edgeR in R.

Transcriptomics Guide →

Single-Cell RNA-Seq (scRNA-seq)

Droplet microfluidics (10x Genomics Chromium) capturing single-cell transcriptomes with UMIs. Cellular heterogeneity clustering and trajectory modeling via Seurat and Scanpy.

Single-Cell Course (13 Ch) →

Spatial Transcriptomics

Preserving the 2D architectural coordinates of gene expression in intact histology tissue sections using 10x Visium, Xenium In Situ, and NanoString CosMx.

Spatial Biology Module →

🔬 3. Quantitative Proteomics: DIA / DDA & Phosphoproteomics

Mass Spec • 3D Structures • PTMs

Proteomics measures the ultimate functional machinery of the cell. Because post-transcriptional attenuation and protein degradation decouple mRNA expression from protein abundance, quantitative liquid chromatography tandem mass spectrometry (LC-MS/MS) provides direct insight into active signaling cascades, post-translational modifications (PTMs), and druggable target complexes.

Standard 5-Stage Architecture

Quantitative Proteomics Pipeline (DIA / DDA & Phosphoproteomics)

📦 Toy Dataset: proteomics_mini_diann_report.tsv (500 Proteins)

STAGE 1

Raw Conversion & Peak Centroiding

Convert proprietary vendor files (.raw, .d) to open indexed centroided .mzML using ThermoRawFileParser or ProteoWizard msconvert.

STAGE 2

Spectral Library & Quantification

DDA (LFQ/TMT): MSFragger + Philosopher + IonQuant.
DIA: Library-free deep learning in-silico spectral library prediction directly from FASTA proteome in DIA-NN.

STAGE 3

QC & Hybrid Imputation

Filter reverse decoys and contaminants (cRAP), enforce ≥2 unique peptides, apply Variance Stabilizing Normalization (VSN), and impute left-censored MNAR values via MinProb / QRILC.

STAGE 4

Differential Abundance

Linear modeling via MSstats or empirical Bayes moderated t-testing in limma / DEP (|log₂FC| > 1, Benjamini-Hochberg FDR < 0.05).

STAGE 5

Phospho & PPI Networks

Kinase-Substrate Enrichment Analysis via KSEAapp for phosphoproteomics and protein-protein interaction (PPI) network hub clustering via STRINGdb / Cytoscape.

Step 01: Raw Vendor File Standardization (ThermoRawFileParser)

# Convert all Thermo .raw files in directory to indexed centroided .mzML (-p = peak picking)
ThermoRawFileParser.sh   -d ./raw_spectra/   -o ./mzml_centroided/   -f 2   -g   -p

Step 02: Library-Free DIA-NN Execution (Two-Pass Deep Learning)

# Step 2A: Generate in-silico predicted spectral library (.predicted.speclib) from FASTA
diann   --lib ""   --threads 32   --verbose 1   --out-lib reference_insilico.speclib   --gen-spec-lib   --predictor   --fasta reference_proteome_canonical.fasta   --fasta-search   --min-fr-mz 200 --max-fr-mz 1800   --min-pep-len 7 --max-pep-len 30   --min-pr-mz 300 --max-pr-mz 1800   --min-pr-charge 1 --max-pr-charge 4   --cut K*,R* --missed-cleavages 1   --unimod4   --var-mods 1 --var-mod UniMod:35,15.994915,M

# Step 2B: Quantify samples against the predicted library with MBR (--reanalyse) & MaxLFQ
diann   --f ./mzml_centroided/ctrl_01.mzML ./mzml_centroided/ctrl_02.mzML ./mzml_centroided/ctrl_03.mzML       ./mzml_centroided/treat_01.mzML ./mzml_centroided/treat_02.mzML ./mzml_centroided/treat_03.mzML   --lib reference_insilico.speclib.predicted.speclib   --threads 32   --out diann_report.tsv   --qvalue 0.01   --matrices   --pg-level 1   --reanalyse   --smart-profiling

Toy Dataset (proteomics_mini_diann_report.tsv)

Protein.Group	Genes	Ctrl_1	Ctrl_2	Ctrl_3	Treat_1	Treat_2	Treat_3	Phospho_Site
P04637	TP53	2450000	2510000	2390000	8900000	9120000	8850000	S315-p
P38398	BRCA1	1820000	1790000	1850000	1950000	1880000	1910000	S1524-p
P00533	EGFR	6200000	6350000	6120000	1200000	1150000	1280000	Y1068-p
P42224	STAT1	3100000	2980000	3250000	7400000	7650000	7300000	Y701-p
P60709	ACTB	45000000	44500000	45800000	45200000	44800000	45100000	Unmodified
P04406	GAPDH	62000000	61800000	62500000	61900000	62200000	62100000	Unmodified

Step 03: VSN Normalization & MinProb Imputation

# R Production Script: VSN Normalization, MinProb Imputation & Limma Differential Abundance
library(DEP)
library(dplyr)
library(SummarizedExperiment)

# 1. Load DIA-NN Protein Group Matrix & filter reverse decoys/contaminants
pg_raw <- read.delim("diann_report.pg_matrix.tsv", stringsAsFactors = FALSE, check.names = FALSE)
pg_clean <- pg_raw %>%
  filter(!grepl("cRAP_|CON__|REV__", Protein.Group)) %>%
  make_unique("Genes", "Protein.Group", delim = ";")

# 2. Experimental design table
exp_design <- data.frame(
  label = c("ctrl_01", "ctrl_02", "ctrl_03", "treat_01", "treat_02", "treat_03"),
  condition = c("Control", "Control", "Control", "Treated", "Treated", "Treated"),
  replicate = c(1, 2, 3, 1, 2, 3)
)

# 3. Construct SummarizedExperiment & filter missing values
sample_cols <- which(colnames(pg_clean) %in% exp_design$label)
se <- make_se(pg_clean, sample_cols, exp_design)
se_filt <- filter_missval(se, thr = 1)

# 4. VSN Normalization & MinProb (q=0.01) left-censored MNAR imputation
se_norm <- normalize_vsn(se_filt)
se_imp <- impute(se_norm, fun = "MinProb", q = 0.01)

# 5. Empirical Bayes testing with limma
se_diff <- test_diff(se_imp, type = "control", control = "Control")
dep_res <- add_rejections(se_diff, alpha = 0.05, lfc = log2(2.0))
write.csv(get_results(dep_res), "proteomics_diff_results.csv", row.names = FALSE)

Step 04: Phosphoproteomics & Kinase Activity Inference (KSEAapp)

# R Production Script: Kinase-Substrate Enrichment Analysis (KSEA)
library(KSEAapp)

# Input format: Protein, Gene, Peptide, Residue.Both (e.g. "S473"), p, FC
px_data <- read.csv("phospho_class1_sites_limma.csv")

# Infer upstream kinase z-scores using PhosphoSitePlus + NetworKIN
KSEA.Complete(
  KSData = KSData,
  PX = px_data,
  NetworKIN = TRUE,
  NetworKIN.cutoff = 5,
  m.cutoff = 5,
  p.cutoff = 0.05
)

🧪 4. Untargeted Metabolomics: XCMS, MZmine 3 & GNPS

Small Molecules • OPLS-DA VIP • Lipids

Metabolomics profiles the downstream metabolic flux resulting from genomic, transcriptomic, and environmental interactions. High-resolution liquid chromatography tandem mass spectrometry (LC-MS/MS) captures thousands of small molecules (<1500 Da), requiring sophisticated peak alignment, signal drift correction across pooled quality controls, and molecular networking.

Standard 5-Stage Architecture

Untargeted LC-MS/MS Metabolomics Pipeline

📦 Toy Dataset: metabolomics_xcms_feature_table.csv (300 Features)

STAGE 1

Centroiding & Peak Picking

Convert vendor files to centroided .mzML via msconvert. Execute XCMS4 CentWave peak detection (ppm = 15, peakwidth = c(5, 30)).

STAGE 2

Alignment & Adduct Deconvolution

Align retention times using Obiwarp, group peak correspondence across sample classes, and deconvolute isotopes/adducts ([M+H]+, [M+Na]+) via CAMERA.

STAGE 3

QC Drift & PQN Scaling

Blank subtraction (Sample/Blank > 3x), pooled QC LOESS drift correction (QC-RLSC), filter features by QC RSD < 20%, and apply Probabilistic Quotient Normalization (PQN) with Pareto scaling.

STAGE 4

Multivariate Chemometrics

Unsupervised PCA (verifying QC cluster tightness) and supervised OPLS-DA with 1,000-permutation testing (R²Y, Q²) in ropls to extract VIP > 1.0 markers.

STAGE 5

MS/MS Matching & Pathways

Structure prediction using SIRIUS 5 + CSI:FingerID, GNPS Feature-Based Molecular Networking (FBMN), and KEGG pathway topological impact enrichment.

Step 01: Peak Detection, RT Alignment & Gap Filling (xcms v4)

# R Production Script: XCMS v4 Peak Picking, Alignment & MS2 Export for GNPS
library(xcms)
library(MsExperiment)
library(Spectra)

# 1. Load metadata & raw files
meta <- read.csv("lcms_injection_sequence.csv")
mse <- readMsExperiment(spectraFiles = meta$file_path, sampleData = meta)

# 2. CentWave Peak Detection
cwp <- CentWaveParam(ppm = 15, peakwidth = c(5, 30), snthresh = 10, prefilter = c(3, 5000), noise = 1000)
xdata <- findChromPeaks(mse, param = cwp)

# 3. Retention Time Alignment (Obiwarp) & Peak Density Correspondence
xdata <- adjustRtime(xdata, param = ObiwarpParam(binSize = 0.6))
pdp <- PeakDensityParam(sampleGroups = sampleData(xdata)$sample_group, minFraction = 0.70, bw = 5, binSize = 0.015)
xdata <- groupChromPeaks(xdata, param = pdp)

# 4. Fill Missing Peaks & Export Feature Matrix
xdata <- fillChromPeaks(xdata, param = ChromPeakAreaParam())
feature_mat <- featureValues(xdata, value = "into", method = "maxint")
feature_def <- as.data.frame(featureDefinitions(xdata))
write.csv(cbind(feature_def[, c("mzmed", "rtmed")], feature_mat), "xcms_raw_feature_table.csv")

Toy Dataset (metabolomics_xcms_feature_table.csv)

Feature_ID,mz,RT_min,Compound,QC_RSD,Ctrl_1,Ctrl_2,Ctrl_3,Dis_1,Dis_2,Dis_3,VIP
FT001,89.0238,1.25,L-Alanine,4.8,48200,47900,49100,51200,50800,51500,0.42
FT002,174.0401,3.42,cis-Aconitate,6.2,124000,121000,126000,38000,36500,39200,2.15
FT003,191.0192,4.81,Citric_Acid,5.1,310000,305000,318000,92000,89000,94500,2.41
FT004,117.0188,2.15,Succinic_Acid,7.4,85000,83000,86500,245000,251000,242000,2.08
FT005,146.0691,5.10,L-Glutamate,8.3,62000,60500,63000,185000,192000,181000,1.89
FT006,203.0531,1.88,D-Glucose,3.9,520000,515000,528000,518000,522000,519000,0.18

Step 02: QC-RLSC Drift & OPLS-DA (pmp + ropls)

library(pmp)
library(ropls)

# 1. Blank Subtraction & Pooled QC LOESS Drift Correction (QC-RLSC)
keep_blk <- rowMeans(feature_mat[, bio_idx], na.rm=TRUE) > (3 * rowMeans(feature_mat[, blank_idx], na.rm=TRUE))
mat_clean <- feature_mat[keep_blk, -blank_idx]
mat_corr  <- QCRSC(df=mat_clean, order=meta_sub$injection_order, batch=rep(1, nrow(meta_sub)), classes=meta_sub$sample_group)
mat_filt  <- filter_peaks_by_rsd(df=mat_corr, classes=meta_sub$sample_group, qc_label="QC", max_rsd=20)

# 2. kNN Imputation & Probabilistic Quotient Normalization (PQN)
mat_imp <- mv_imputation(mat_filt, method="knn")
mat_pqn <- pqn_normalisation(mat_imp, classes=meta_sub$sample_group, qc_label="QC")

# 3. Supervised OPLS-DA (Pareto Scaling, 1000 Permutations)
X_bio <- t(log2(mat_pqn[, bio_mask] + 1))
Y_bio <- factor(meta_sub$sample_group[bio_mask])
opls_mod <- opls(x=X_bio, y=Y_bio, predI=1, orthoI=NA, scaleC="pareto", permI=1000)
vip_scores <- getVipVn(opls_mod)
write.csv(data.frame(Feature=names(vip_scores), VIP=vip_scores), "oplsda_vip_features.csv")

Step 03: In-Silico Structure Annotation (SIRIUS 5 + CSI:FingerID CLI)

# Run SIRIUS + ZODIAC + CSI:FingerID + CANOPUS on exported MS2 .mgf spectra
sirius   --input xcms_ms2_features.mgf   --output ./sirius_project/   --max-mz 850   --cores 16   formula -p orbitrap --ppm-max 5 --candidates 10   zodiac   fingerid --db BIO   canopus   write-summaries --output ./sirius_summaries/

🧬 5. Epigenomics: WGBS/EM-Seq, ATAC-Seq & CUT&Tag

DNA Methylation • Chromatin • Footprinting

Epigenomics reveals how identical genomic sequences produce divergent cellular phenotypes. By profiling 5-methylcytosine (5mC) CpG methylation, Tn5 transposase-accessible chromatin, and histone post-translational modifications, epigenomics uncovers active promoters, distant enhancers, and transcription factor occupancy across development and disease.

Standard 5-Stage Architecture

Epigenomics Pipeline (WGBS/EM-Seq Methylation, ATAC-Seq & CUT&Tag)

📦 Toy Dataset: chr1_mini_bismark.cov.gz & atac_mini_peaks.narrowPeak

STAGE 1

Trimming & Conversion QC

fastp adapter trimming and Bismark 3-letter alignment. Verify C-to-T non-conversion rate on unmethylated lambda phage or chloroplast spike-in controls (>99% conversion efficiency required).

STAGE 2

Context Methylation & DMRs

Extract context-specific methylation (CpG, CHG, CHH) via bismark_methylation_extractor and call Differentially Methylated Regions (DMRs) using DSS beta-binomial models.

STAGE 3

Chromatin Accessibility

ATAC-Seq: Shift Tn5 cut sites (+4 bp / -5 bp via alignmentSieve --ATACshift), filter fragments <100 bp, and call peaks with MACS3 --nomodel --shift -100 --extsize 200.
CUT&Tag: Sparse peak calling with SEACR normalized to IgG control.

STAGE 4

Feature & Motif Footprinting

Annotate peak distribution relative to TSS (±3 kb) using ChIPseeker, generate TSS/TES matrix profiles with deepTools, and scan TF footprints with TOBIAS and HOMER findMotifsGenome.pl.

STAGE 5

Chromatin-to-Expression

Multi-omics integration: Correlate promoter DMR hypomethylation and ATAC-Seq open chromatin peak gains with downstream gene upregulation in paired DESeq2 RNA-Seq datasets.

Track A: WGBS & EM-Seq Methylome Pipeline (Bismark + DSS)

# 1. Prepare Bisulfite/EM-Seq Reference Index & Align Paired-End Reads
bismark_genome_preparation --bowtie2 --parallel 8 ./ref_genome_dir/
bismark --genome ./ref_genome_dir/ -1 R1.fq.gz -2 R2.fq.gz --parallel 4 -p 4 --output_dir ./bismark_aligned/
deduplicate_bismark --bam -p ./bismark_aligned/sample_bismark_bt2_pe.bam

# 2. Extract CpG/CHG/CHH Context Methylation Ratios
bismark_methylation_extractor --paired-end --no_overlap --comprehensive --bedGraph --CX_context   --cytosine_report --genome_folder ./ref_genome_dir/ --parallel 8   sample_bismark_bt2_pe.deduplicated.bam

Track B: ATAC-Seq Tn5 Shift, MACS3 Peaks & TOBIAS Footprinting

# 1. Align with Bowtie2, remove chrM, and filter MAPQ >= 30
bowtie2 --very-sensitive -X 2000 -p 16 -x ref_index -1 atac_R1.fq.gz -2 atac_R2.fq.gz |   samtools view -b -u -q 30 -F 1804 -f 2 - |   samtools idxstats - | cut -f 1 | grep -v -E 'chrM|chrPt|Mt|Pt' |   xargs samtools view -b raw.bam > atac_nuclear_filtered.bam

# 2. Shift Reads for Tn5 Cut-Site (+4 bp / -5 bp) & Call Peaks with MACS3
alignmentSieve --numberOfProcessors 16 --ATACshift --maxFragmentLength 100 --bam dedup.bam -o atac_NFR_shifted.bam
macs3 callpeak -t atac_NFR_shifted.bam -f BAM -g 2.7e9 -n sample_atac --nomodel --shift -100 --extsize 200 --keep-dup all -q 0.01

# 3. Bias-Corrected Transcription Factor Footprinting with TOBIAS
TOBIAS ATACorrect --bam atac_NFR_shifted.bam --genome ref.fa --peaks consensus_peaks.bed --outdir tobias_out --cores 16
TOBIAS ScoreBigwig --signal tobias_out/atac_NFR_shifted_corrected.bw --regions consensus_peaks.bed --output footprints.bw --cores 16
TOBIAS BINDetect --motifs JASPAR_motifs.meme --signals footprints.bw --genome ref.fa --peaks consensus_peaks.bed --outdir bindetect_out --cores 16

Track C: CUT&Tag E. coli Spike-In Scaling & SEACR Sparse Peak Calling

# 1. Scale fragments by E. coli spike-in factor (10,000 / ecoli_reads)
SCALE_FACTOR=$(awk -v ecoli_reads=12500 'BEGIN {print 10000 / ecoli_reads}')
bedtools bamtobed -bedpe -i cutandtag_namesorted.bam | awk '$1==$4 && $6-$2 < 700 {print $1"\t"$2"\t"$6}' | sort -k1,1 -k2,2n > fragments.bed
bedtools genomecov -bg -scale "$SCALE_FACTOR" -i fragments.bed -g ref.chrom.sizes > sample_norm.bedgraph

# 2. Call Stringent Enriched Domains Against IgG Control with SEACR
bash SEACR_1.3.sh sample_norm.bedgraph igg_control_norm.bedgraph non stringent sample_cutandtag_seacr

Toy Dataset (atac_mini_peaks.narrowPeak)

chr1	1002500	1003100	ATAC_Peak_001	450	.	8.42	12.5	9.8	320
chr1	1245000	1245800	ATAC_Peak_002	890	.	15.10	24.8	18.2	410
chr2	3205000	3205900	ATAC_Peak_003	320	.	5.80	8.1	6.4	510
chr3	8420000	8420700	ATAC_Peak_004	920	.	18.25	32.4	26.1	390
chr7	5515000	5516200	ATAC_Peak_005	1100	.	22.40	45.1	38.5	620
chr11	65280000	65280800	ATAC_Peak_006	780	.	12.90	19.5	15.4	400

Executable R ChIPseeker & DSS Recipe

library(ChIPseeker)
library(TxDb.Hsapiens.UCSC.hg38.knownGene)
txdb <- TxDb.Hsapiens.UCSC.hg38.knownGene

# 1. Load atac_mini_peaks.narrowPeak & annotate relative to TSS (+/- 3kb)
peaks <- readPeakFile("atac_mini_peaks.narrowPeak")
peakAnno <- annotatePeak(peaks, tssRegion=c(-3000, 3000), TxDb=txdb)
print(as.data.frame(peakAnno)[, c("seqnames", "start", "end", "annotation", "distanceToTSS")])

# 2. Beta-Binomial DMR Calling with DSS
library(DSS)
# dml_test <- DMLtest(BSobj, group1 = c("C1", "C2"), group2 = c("T1", "T2"), smoothing = TRUE)
# dmrs <- callDMR(dml_test, delta = 0.20, p.threshold = 1e-4)

🌱 6. Plant, Polyploid & Non-Model Genomics: Pangenomes, TEs & Crop GWAS

Wheat • EDTA • SubPhaser • Minigraph-Cactus

Agricultural crops and wild non-model plants represent the frontier of computational biology complexity. Species like allohexaploid bread wheat (Triticum aestivum, 16 Gb), polyploid sugarcane (10 Gb), and octoploid strawberry possess multi-subgenome architectures, extreme repeat densities (>80% retrotransposons), and rampant homoeologous exchange that break standard mammalian analysis pipelines.

🧬

Polyploid Phasing & Subgenomes

Resolve subgenome assemblies (A, B, D) without ancestral references using SubPhaser k-mer enrichment, PacBio HiFi string graphs in Hifiasm, and MCScan collinear synteny mapping.

📦

EDTA De Novo TE Annotation

Structure-based identification of Ty1/Copia, Ty3/Gypsy, Helitrons, and TIR transposons using the EDTA pipeline, with LTR Assembly Index (LAI) validation to certify telomere-to-telomere continuity.

🌾

Crop Graph Pan-Genomics & GWAS

Capture dispensable crop resistance genes with Minigraph-Cactus variation graphs, and map complex quantitative traits with GEMMA mixed models accounting for breeding population structure.

Full Plant Track (13 Chapters) →

SYSTEMS BIOLOGY FUSION

🌐 7. Multi-Omics Data Integration & Machine Learning

Single-omics analyses only capture isolated slices of biological reality. Modern precision medicine increasingly relies on Multi-Omics Data Fusion—combining genomics (germline/somatic mutations), transcriptomics (expression), proteomics, and metabolomics into integrated statistical models.

Figure 1.7: Multi-Omics Factor Analysis Integration Architecture — Unsupervised Bayesian matrix decomposition combining orthogonal molecular layers into latent factors for biomarker discovery and clinical outcome prediction.

Unsupervised Matrix Factorization (MOFA+)

Multi-Omics Factor Analysis identifies latent dimensions of variation that explain shared vs. assay-specific biological variance across heterogeneous sample cohorts.

Single-Cell Multi-Modal Assays

Assaying multiple molecular layers simultaneously from the exact same single cell: CITE-seq (RNA + cell-surface antibodies), 10x Multiome (RNA + ATAC accessibility).

Network Biology & Graph Neural Networks

Overlaying multi-omics signatures onto biological knowledge graphs (STRING protein-protein interactions, KEGG metabolic reactions) to discover robust clinical biomarkers.

📋 Master Omics Disciplines Comparison Matrix

Quick Reference

Omics Discipline	Target Molecule	Primary Technology	Core Data Formats	Standard Software Tools	Global Public Repository
Genomics	Nuclear / Mt DNA	Illumina, PacBio HiFi, ONT	FASTQ, BAM, VCF	BWA-MEM2, GATK4, DeepVariant	NCBI SRA, ENA
Transcriptomics	mRNA, lncRNA, miRNA	RNA-Seq, 10x Chromium, Visium	FASTQ, Counts, MTX	STAR, Salmon, DESeq2, Seurat	NCBI GEO, ArrayExpress
Proteomics	Proteins, Peptides, PTMs	LC-MS/MS, Cryo-EM, X-ray	RAW, mzML, PDB	MaxQuant, DIA-NN, AlphaFold 3	PRIDE, ProteomeXchange, PDB
Metabolomics	Metabolites (<1.5 kDa)	LC-MS, GC-MS, NMR	mzML, netCDF, Peak tables	XCMS, MetaboAnalyst, GNPS	MetaboLights, Metabolomics Workbench
Epigenomics	5mC DNA, Histones, Open Chromatin	WGBS, ATAC-seq, ChIP-seq	BAM, BigWig, narrowPeak	MACS3, Bismark, chromHMM	ENCODE Portal, GEO
Metagenomics	Microbial Community DNA	16S Amplicon, Shotgun WMS	FASTQ, BIOM, FASTA (MAGs)	DADA2, QIIME 2, Kraken 2, CheckM	MG-RAST, NCBI SRA, ENA