Codon dwell-time tables (per-codon timing)¶

Per-codon timing (Synthesis on a coarse-grained ribosome) needs a codon dwell-time table — the codon_times key of a csp.ini / cylinder.ini. It maps each codon to its mean in-vivo per-codon translation (dwell) time in seconds (the codon’s mean first-passage time), which is what sets each residue’s MD-segment length. There is no bundled default: you pick a table for your organism (and, where relevant, elongation rate). cosmo ships a small library of them.

Where they live & the format¶

The tables are under assets/csp/codon_dwell_times/<organism>/. Each is a plain text file, one row per codon:

# codon   dwell_time_seconds   amino_acid
UUU       0.068164             PHE
...
UAA       0.005513             STOP

RNA alphabet (U), tab-separated, # comments ignored; 64 rows (61 sense + 3 stop). Larger time = slower codon; stop codons carry the amino acid STOP.
The third column (amino acid) is what mrna = fastest/slowest/median uses to pick each residue’s synonymous codon — see Fastest / slowest / median mRNA.
Every organism directory also has a *_dwell_time_methods.md with the full provenance (source dataset, references, how relative data were scaled to seconds).

Shipped tables¶

Organism	File(s)	Basis (see the methods note for full provenance)
E. coli	`ecoli_codon_dwell_times_310K.txt`	Fluitt, Pienaar & Viljoen (2007) mechanistic decoding times, O’Brien-rescaled so the codon-usage-weighted mean = 0.061 s/codon (16.5 aa/s) at 310 K. Absolute; the most-validated table.
S. cerevisiae	`yeast_codon_dwell_times.txt`	Gardin et al. (2014) relative decoding rates (RRT), frequency-weighted → seconds (~0.108 s/codon; ~9.3 codons/s).
	`yeast_codon_dwell_times_OBrien.txt`	O’Brien-lab chemical-kinetic model, absolute (~0.233 s/codon; ~4.3 codons/s) — the most pipeline-native yeast table.
	`yeast_codon_dwell_times_OBrien_rescaled_9.3.txt`	The O’Brien table rescaled to 9.3 aa/s (matches the Gardin rate).
H. sapiens	`human_codon_dwell_times.txt`	Gobet et al. (2020) liver ribosome-profiling dwell times (Lintner dataset) → seconds.
	`human_codon_dwell_times_v2_freqweighted.txt`	Same, frequency-weighted to an elongation-rate anchor R = 5.6 codons/s.
N. crassa	`ncrassa_codon_dwell_times_R{5.35_low,6.7_mid,8.02_high}.txt`	Yang et al. (2019) relative codon decoding times (RCDT) → seconds at three measured elongation-rate anchors (5.35 / 6.7 / 8.02 codons/s). Least-certain table (relative values digitized from a figure).
	`..._v2_freqweighted.txt` (× 3)	Frequency-weighted variants of the three anchors.

Frequency-weighted (v2) variants. These rescale the per-codon times so the codon-usage-weighted mean over the organism’s transcriptome hits the target elongation rate, instead of an unweighted mean. Use them when you want the genome-wide average speed to match the anchor; use the plain (v1) tables for the raw per-codon values.

Note

References — E. coli: Fluitt, Pienaar & Viljoen, Comput. Biol. Chem. 31:335–346 (2007). Yeast: Gardin et al., eLife 3:e03735 (2014). Human: Gobet et al., PNAS 117(17):9630–9641 (2020). N. crassa: Yang et al., Nucleic Acids Res. 47(17):9243–9258 (2019).

Use one in a run¶

Point the codon_times key at a table (a path selects per-codon timing; a plain number would instead mean a uniform time for every codon):

pdb_file    = my_protein.pdb
mrna        = my_protein_mrna.txt
codon_times = assets/csp/codon_dwell_times/ecoli/ecoli_codon_dwell_times_310K.txt

Keep the thermostat consistent with the table’s temperature (the E. coli table is 310 K); set ref_t to your table’s temperature. See Synthesis control options for the full key reference.

Fastest / slowest / median mRNA¶

Because each table carries the codon→amino-acid mapping (column 3), you can have cosmo build an mRNA that keeps the protein sequence fixed but reassigns each residue’s codon — a controlled walk along the synonymous-mutation (codon-optimization) axis. Since the protein, its native structure, and its contacts are identical across these mRNAs and only the elongation timing changes, any difference in the co-translational folding they produce is attributable to codon kinetics alone. Set mrna to a keyword instead of a filename:

fastest — the shortest-τ synonymous codon at every residue → the fastest translation the protein’s codons allow (minimises the mean total dwell time).
slowest — the longest-τ synonymous codon at every residue → the slowest translation (maximises the mean total dwell); the slow codons act as translational pauses that give each emerging segment more time to fold.
median — the middle-τ synonymous codon at every residue → a neutral reference between the two extremes. When an amino acid has an even number of synonymous codons there is no single middle, so cosmo takes the faster (shorter-τ) of the two central codons; the pick is deterministic.

Here τ is the codon’s mean per-codon dwell time from the chosen codon_times table. The runner reads the protein sequence from pdb_file, picks one codon per residue as above, appends a terminating stop codon, and writes mrna_fastest.txt / mrna_slowest.txt / mrna_median.txt next to the PDB. Because the pick is made per amino acid, the mode needs a codon_times table (it defines which codon is fast/slow/median); a uniform numeric codon_times is rejected. To pre-generate one standalone:

cosmo-make-mrna --pdb my_protein.pdb \
    --codon-times assets/csp/codon_dwell_times/yeast/yeast_codon_dwell_times_OBrien.txt \
    --mode slowest