# DNA synthesis and gene foundries

The phosphoramidite cycle and why it caps oligo length, how errors are found and corrected, what enzymatic synthesis changes, and why cumulative yield rules the design space.

Writing DNA means adding one base at a time to a chain that only compounds losses: per-step yield, per-base errors and cost per base, not imagination, set what can be designed.

Source: https://en.bioecon.ru/docs/synbio-enabling/foundries-design/dna-synthesis-gene-foundry/
Updated: 2026-09-07



Genes are not written in one piece. Every synthesiser builds DNA the same way — one base per cycle onto the end of a growing chain — and everything about the industry follows from the arithmetic of that loop: a step that is only almost perfect, repeated two hundred times.

## One base at a time

Phosphoramidite synthesis is a four-step cycle on solid support: acid removes the protecting group at the chain end (detritylation), the next activated base couples, any chain that failed to extend is capped so it can never catch up, and the new linkage is oxidised to a stable phosphate. The chemistry is fast and has run essentially unchanged for decades. Its limit is statistical. Coupling succeeds a fraction of a percent short of perfection, and the shortfall compounds: the fraction of chains with every base in place falls geometrically with length, so a long crude oligo is dominated by truncated chains. The acid detritylation step also slowly depurinates the growing strand, which later breaks it. This is why routine oligos stop around two hundred nucleotides — not because the machine cannot physically continue, but because the full-length product is being taxed exponentially.

## Errors and correction

A wrong base is rare per position, but a gene is thousands of positions, so a naive assembly carries errors in most clones and verification by sequencing is mandatory, not optional. Correction comes in two honest forms. Consensus methods sequence several independent clones and build the majority sequence — reliable, priced in sequencing reads. Enzymatic error correction exploits the double helix: strands are re-annealed so mismatches form physical bulges, and a mismatch-binding protein (a MutS homologue) or a mismatch-cutting nuclease removes the disagreeing segments, leaving a pool enriched for the consensus before re-amplification. Both methods vote; neither creates information, so correction still needs enough correct molecules to vote from. Enzymatic synthesis with terminal deoxynucleotidyl transferase runs in water with no acid step and could in principle write longer, cleaner oligos — but the polymerase adds bases without a template and does not naturally stop after one, so single-base control needs engineered reversible terminators. That problem, not enthusiasm, sets its pace.

## What length and cost actually rule

Genes are assembled from overlapping oligos: annealed overlaps seed polymerase to fill the gaps, and errors from the oligo pool carry through into the assembly. So the design space is set by two coupled quantities. Length per write is capped by cumulative yield; the number of writes is capped by cost per base, which sets whether a project is one gene, a whole pathway, or a library of thousands of variants. Codon optimisation, DNA-level tuning for expression in [a production host](../metabolic-engineering/), is cheap in software and priced in synthesis. The design file that a [CAD platform](../synthetic-biology-cad-software/) emits is therefore never a free-form wish: it is a compromise between what the biology wants, what the chemistry can write and what the budget tolerates.