emusynth 0.1.0

Host-side streamer for the emusynth ABI 3400 solenoid emulator

emusynth -- Notes / parking lot

Things we've decided to defer but should come back to soon.

Next target devices to model (after the 3400 baseline works)

The 3400 is a fine prototype for control-system architecture but caps at ~200 nt phosphoramidite chemistry. To exercise the design at higher fidelity toward "modern long-read synthesizer" use cases, the next manuals/specs to chase:

Tier 1: realistic, public docs, similar architecture (still phosphoramidite)

• BioAutomation MerMade 192 / 384. High-throughput phosphoramidite, 192 or 384 parallel wells, large valve array. Manuals circulate on used-equipment sites and in core-facility documentation. Same chemistry as the 3400 (still ~200 nt cap) but the valve multiplexing is much wider -- forces the decoder to handle a real well-select space, not just 4 columns. Best architectural step up.
• PerSeptive / ABI Expedite 8909. Sibling of the 3400, abundant docs, well understood. Useful as a sanity-check / portability target for the existing procedure tables.
• GE/Cytiva AKTA Oligopilot. Gram-scale phosphoramidite, packed-bed reactor (different topology). Manuals public. Useful if we want to model manufacturing-scale rather than research-scale.

Tier 2: real long-read targets, harder to access

• DNA Script SYNTAX. Benchtop enzymatic (TdT), commercial since ~2021, currently ~80 nt and extending. User manual gated behind customer support. Workaround: DNA Script's patent portfolio describes their fluidics in block-diagram detail. Search Google Patents for assignee="DNA Script" with terms like "fluidic", "reactor", "synthesis cycle". Patents are nearly as good as a manual for our prototyping needs.
• Ansa Biotechnologies. Service-only (no benchtop). Same patent strategy.
• Molecular Assemblies. Service-only. Same patent strategy.

Tier 3: academic / open

• Palluk et al. 2018 (Nature Biotechnology) -- foundational TdT enzymatic synthesis paper. Describes chemistry + reactor in enough detail to write a "v3 enzymatic" procedure table.
• DIY synbio community projects (Hackaday, GitHub). Variable quality, none are kilobase-class.

Recommended sequence of next iterations

1. Finish 3400 model (where we are now): LEDs, Pi decoder, end-to-end test.
2. MerMade 192 model: same chemistry, much wider valve multiplexing. Stress-tests the decoder's ability to handle many parallel wells.
3. Enzymatic v3 procedure table: built from DNA Script + Ansa patents plus the Palluk paper. Demonstrates chemistry-agnostic generality of the architecture (same Uno, same Pi decoder, just different Step table).
4. Decision point: continue refining the emulator, or start designing real hardware for one of the above chemistries.

Other deferred items

• Multi-column parallel synthesis on the 3400 (interleaved column-select). Currently v1 runs one column per RUN/STREAM.
• Modified bases (positions 5-8, valves V27-30) for labeled / non-canonical nucleotides.
• RNA and phosphorothioate cycle scripts (LV200-RNA, LV200-PS).
• Standalone operation (no host computer) -- requires moving from Uno to a board with more SRAM (Mega 2560, Teensy 4.x, ESP32) so the full sequence can live on-instrument again. Streaming protocol stays as the primary path.
• v2 hardware: 24V realistic solenoid drive + PC817 opto level-shift to Pi. Sketched in WIRING.md, not yet built.