2: Experimental Design, Library Prep & Coverage Estimation
Chapter 2: Experimental Design, Library Prep & Coverage Estimation
A rigorous genomic experiment begins long before sequencer run initiation. Insufficient read depth results in false negatives in variant discovery, while excessive sequencing wastes project grant funds. Calculating the theoretical and operational coverage required for specific biological questions is the foundational discipline of practical genomics.
The theoretical average sequencing depth (\(C\)) over a target genome or exome of size \(G\) is defined by the classical Lander-Waterman formula:
C = (N × L) / G
Where N is the total number of sequencing reads, L is the read length in base pairs, and G is the haploid genome or target region size in base pairs (for humans, \(G pprox 3.1 imes 10^9\) bp).
1. Poisson Sampling and Realized Physical Coverage
Sequencing is an unbiased random sampling process governed by a Poisson distribution. The probability that any given nucleotide in the genome is sequenced exactly \(k\) times when the average depth is \(C\) is:
P(k) = (Ck × e-C) / k!
Consequently, the fraction of the genome left completely unsequenced (\(k = 0\)) is \(P(0) = e^{-C}\). At an average depth of \(10 imes\), \(e^{-10} pprox 0.000045\) (0.0045% unsequenced). However, due to GC content variation, secondary structure, and mappability issues, practical clinical guidelines dictate a minimum of 30× mean depth for germline WGS and 80–100× for somatic oncology tumor samples.
2. Library Preparation Chemistries
Constructing a high-quality sequencing library requires fragmenting high-molecular-weight (HMW) genomic DNA and appending platform-specific adapter sequences:
- Mechanical Acoustic Shearing (Covaris): Uses focused acoustics to fragment DNA into random 350–500 bp fragments without sequence-specific cleavage bias. Ideal for PCR-free WGS.
- Enzymatic Tagmentation (Nextera / Illumina DNA Prep): Utilizes hyperactive engineered Tn5 transposases loaded with sequencing adapters. In a single simultaneous step, Tn5 fragments DNA and appends adapters ("tagmentation"). Extremely rapid, but exhibits subtle sequence insertion motifs at GC extremes.
- PCR-Free Workflows: Eliminates library amplification cycles. Essential for human whole-genome sequencing because PCR cycles introduce GC bias (dropout of AT-rich and GC-rich promoters) and duplicate reads.
3. Practical Throughput & Multiplexing Calculation
# Scenario: Sequence 24 human tumor-normal pairs (48 samples total)
# Target: 60x depth for Tumors, 30x depth for Normals
# Human genome: 3.1 Gb (3.1 x 10^9 bp)
# Total required bases:
# 24 Tumors @ 60x = 24 * (60 * 3.1 Gb) = 4,464 Gb
# 24 Normals @ 30x = 24 * (30 * 3.1 Gb) = 2,232 Gb
# Total yield needed = 6,696 Gb (~6.7 Terabases)
# Flow cell selection:
# NovaSeq X 25B Flow Cell yields ~8 Terabases (2 x 150 bp)
# Recommendation: Pool all 48 unique dual-indexed (UDI) libraries onto one 25B flow cell.Common Pitfalls & Troubleshooting Checklist
- Java Heap Exhaustion (GATK / Picard): If processes terminate with
java.lang.OutOfMemoryError, set your environment variable explicitly:export _JAVA_OPTIONS="-Xmx16g -Djava.io.tmpdir=/tmp". - Index Timestamp Mismatch: If tools complain that a BAM index is missing or out of date, regenerate with
samtools index -@ 4 sample.bam. - Conda Channel Conflicts: Avoid ABI incompatibilities between bioconda and conda-forge by setting strict priority:
conda config --set channel_priority strict. - Chromosome Naming Contamination: Always verify whether reference FASTA headers use
chr1or1; divergent naming causes silent alignment drops.