Adaptive Trimming

The 3t-seq pipeline implements an Adaptive Trimming mechanism that automatically optimizes Trimmomatic parameters based on the specific quality profile of your sequencing data.

Motivation

Standard fixed-threshold trimming can sometimes be too aggressive (losing valuable data) or too lenient (leaving artifacts that impact alignment). Adaptive trimming bridges this gap by analyzing FastQC results before executing the trim step.

How it Works

The adaptive trimming process consists of three main phases:

1. Quality Profiling

The pipeline first runs a "raw" FastQC pass on the input FASTQ files. The results are parsed to extract:

• Adapter Content: Detection of Illumina Universal, Nextera, or Small RNA adapters.
• Per-base Quality: Specifically targeting the \(10^{th}\) percentile (\(p_{10}\)) quality score. In simpler terms, the pipeline looks at the "worst 10%" of your sequence data to gauge how poor the quality gets, which helps it decide how aggressive the trimming needs to be.
• Read Length: The maximum observed read length in the library.

2. Parameter Derivation

Using the extracted statistics, the pipeline dynamically assembles a Trimmomatic command:

Feature	Logic	Default / Derived
Adapter Clipping	Detects the most prevalent adapter type to ensure all synthetic sequences are removed.	TruSeq3 or NexteraPE
Quality Trimming	Uses standard window trimming by default, but switches to a more aggressive, information-preserving method (MAXINFO) if the worst 10% of data has a quality score below 15.	Switches to MAXINFO:40:0.5 if average \(p_{10} < 15\).
Minimum Length	Calculates a safe cutoff for downstream alignment.	\(\max(36, 0.35 \times \text{max_length})\)
Base Trimming	Ensures low-quality ends are removed.	LEADING:3 and TRAILING:3

3. Execution

The derived parameters are saved as a JSON metadata file and passed to Trimmomatic. This ensure that each sample in your experiment receives the optimal level of processing.

---

User Overrides

While the system is designed to be fully automated, you can still provide fixed parameters in your config.yaml:

sequencing_libraries:
  - name: MySample
    trimmomatic:
      adaptive: true
      extra_params: "MINLEN:50 CROP:100"

Precedence

User-provided modules in extra_params always take precedence over derived ones. For example, if you provide MINLEN:50, the adaptive MINLEN calculation will be skipped for that sample.

When to Use It

Adaptive trimming is enabled by default. It is particularly recommended for:

• Samples with variable quality across batches.
• Pilot studies where adapter content or quality decay is unknown.
• Large-scale integrative analyses of public datasets from different sources.