📄 / / profiles / r-and-d / raw_evidence / OpenAI4S / skills / bioskills / bio-sequence-manipulation-codon-usage / references / usage-guide.md

Codon Usage - Usage Guide

Overview

This skill enables AI agents to help you analyze codon usage in coding sequences using Biopython: codon counting, CAI (Codon Adaptation Index) scoring against a host, RSCU and Nc bias metrics, GC at codon positions, and naive max-CAI codon optimization for heterologous expression. It uses the redesigned Bio.SeqUtils.CodonAdaptationIndex class (the old Bio.SeqUtils.CodonUsage module was removed in Biopython 1.82).

Prerequisites

pip install biopython

Quick Start

Tell your AI agent what you want to do:

"Calculate codon frequencies and RSCU for this coding sequence"
"What is the CAI of this gene against my E. coli reference genes?"
"Build a CAI index from these highly expressed genes and score my construct"
"Codon-optimize this gene for the host, but warn me about expression risks"
"Find rare/low-RSCU codons in my sequence"
"What is the GC content at each codon position?"

Example Prompts

CAI Against a Host Reference

"Build a Codon Adaptation Index from this FASTA of highly expressed E. coli genes and report the CAI of my target gene."

Codon Optimization

"Codon-optimize this CDS for the host using max-CAI, confirm the protein is unchanged, and list which tradeoffs (ramp, 5' structure, cryptic sites) I should screen before ordering it."

RSCU Analysis

"Calculate RSCU values for my coding sequence and flag over- and under-used synonymous codons."

Rare Codon Detection

"Find the codons with RSCU below 0.5 in this gene."

Overall Bias

"Compute the effective number of codons (Nc) for this sequence."

Wobble Position Bias

"What is the GC content at the first, second, and third codon positions?"

Mitochondrial or Bacterial Genes

"Score this mitochondrial gene's CAI using the vertebrate mitochondrial codon table."

What the Agent Will Do

Import CodonAdaptationIndex from Bio.SeqUtils (not the removed CodonUsage module)
For CAI: build the index from a reference set of highly expressed host genes, then call calculate
Confirm the query is in frame before scoring (out-of-frame input is silently mis-scored)
For optimization: call optimize, verify the protein is preserved, and flag expression-risk tradeoffs
Compute requested bias metrics (RSCU, Nc, GC123) and return formatted results

Key Concepts

CAI (Codon Adaptation Index)

Geometric mean of per-codon relative adaptiveness (w), range 0-1, higher = better adapted
w must come from highly expressed genes of the TARGET organism; CAI against a whole-genome or wrong-organism reference is meaningless
ATG, TGG, and stop codons are excluded; unobserved codons get w = 0.5

RSCU (Relative Synonymous Codon Usage)

Observed count / expected-if-uniform within a synonymous family
1.0 = no bias, >1 over-used, <1 under-used; the w weights in CAI are built from RSCU ratios

Nc (Effective Number of Codons)

Reference-free overall bias measure, range ~20 (fully biased) to 61 (unbiased)

tAI (tRNA Adaptation Index)

Supply-side alternative weighting codons by tRNA gene copy number and wobble efficiency
Often tracks expression better than CAI, but is not in Biopython (use the R tAI package)

Why naive max-CAI optimization can hurt expression

optimize() only maximizes single-codon CAI; it ignores the translation ramp, 5' mRNA structure, GC extremes, cryptic regulatory elements, and codon-pair bias
Failure is silent: correct protein, high CAI, poor expression. Treat the output as a draft to screen.

Tips

The biggest trap: from Bio.SeqUtils import CodonUsage raises ImportError on Biopython >= 1.82. Use from Bio.SeqUtils import CodonAdaptationIndex.
There is no bundled Sharp E. coli index. Always build the index from your own reference CDS.
Keep queries in frame (length divisible by 3, first base = codon 1); out-of-frame input does not raise, it is silently corrupted.
Pass the matching table= for mitochondrial or bacterial genes.
optimize(strict=True) raises on a tie; use strict=False to let Biopython pick one.
GC123 returns percentages (0-100); gc_fraction returns a 0-1 fraction. GC3 correlates with genome GC.

transcription-translation - Translate CDS and select the correct codon table
sequence-properties - GC123 and per-position GC content
sequence-io/read-sequences - Parse reference CDS from FASTA/GenBank for CAI training
database-access/entrez-fetch - Fetch highly expressed gene sets from NCBI for CAI references