jvrana / dasi-dna-design

A Python molecular biology CAD tool with a dead-simple user interface

DASi DNA Design

DASi is an automatic DNA cloning plan designer aimed for operating on small budgets by focusing on material re-use.

The software converts a nucleotide sequence, or a library of sequences, to an executable molecular assembly plan while optimizing material cost, assembly efficiency, and assembly time.

The key design paradigm for DASi is that no molecular biology expertise is required to use DASi. Complete novices should be able to use the software to design and construct new genetic sequences. This also enables automated software programs to automatically design and construct new genetic sequences.

The software goals are reminiscent of j5 or Teselegen but focused on:

1. A dead-simple API usable by lab novices, experts or automated software programs.
2. Utilizing information about current laboratory inventory in its optimization algorithm to minimize costs and turn-around time

Status

DASi is currently under development funded by the DARPA Synergistic Discovery and Design program. DASi is currently being used to connect automatically generate DNA designs to automated biological fabrication facilities (e.g. University of Washington Biofab).

Usage

DASi completely automates the cloning design work, finding approximately optimal solutions for cloning steps, preferentially using existing plasmids, linear DNA fragments, and primers to design semi-optimal cloning steps and designs.

The following command designs the cloning steps for a library of designs. The user only needs to specify the sequences they wish to construct and currently available primers and DNA templates as .genbank or .fasta files. DASi handles all design aspects. No molecular biology expertise is required to use DASi.

dasi library_design --designs mydesigns/*.gb --fragments fragments/*.gb --primers primers.fasta --templates plasmids/*.gb --cost_model cost.b --out results

Customization

DASi optimization parameters are completely customizable. The following are examples of parameters and aspects of DASi that are customizable:

• primer synthesis costs
• primer design parameters
• synthetic fragment costs
• vendor-specific synthetic fragment complexity
• sequence dependent plasmid assembly efficiencies
• optimizing over efficiency vs material costs
• etc.

Planned Features

• Golden-gate support
• heirarchical assembly
• library support (with bayesian search to optimize shared parts)
• front-end
• connection to fabrication facility

DASi optimization problem

Briefly, DASi approximates a solution the following optimization problem:

Given a set of 'goal' double-stranded sequences, a set of available single-stranded and double-strand sequences, and a set of actions that can create new sequences, find the optimal set of operations that produces the 'goal' sequences.

Formalization of this optimization problem is coming soon.

dasi-dna-design's Issues

RuntimeWarning: invalid value encountered in true_divide

Describe the bug
The following warning often occurs in the normal execution of DASi
RuntimeWarning: invalid value encountered in true_divide

To Reproduce
Run DASi on any sequence design

Expected behavior
No warnings

Additional context
This occurs in the networkx utilities algorithms.

Vendor Configurations (e.g. IDT, Twist)

Is your feature request related to a problem? Please describe.
There should be default vendor configurations for costs.

Describe the solution you'd like
Vendor considerations
IDT

• Well defined length > cost and time estimates

TWIST

• Produces adapter on each fragment. Primers must be ordered for assemblies in most cases. This optimization procedure is not implemented. For TWIST synthesized fragments, the cost would be equivalent to the synthesis cost + the cost of new primers. There may be optimizations to be had with ordering the TWIST fragment and ordering different primers to amplify the fragment.

Testing Utilities: More control over random library generation

• Creating designs that have sequence similarity
• Creating designs that use pre-prepared DNA fragments
• Creating designs that utilize pre-made primers

Information about input sequences should be present in

Optimal Partitioning Schema

Milestones

1. Implement DNA complexity to efficiency for graphs
2. Implement partitioning for B->A gaps
3. Implement partitioning algorithm to find optimal synthesis partitioning

Understand why some designs fail to assemble

Input parameters should be exposed

There are a lot of design parameters, these should be exposed to the user.

Randomized tests occasionally fails

test_fake_design

JSON output schema

DASi should produce a clear, easy to understand JSON output for its designs.

speed improvements

Linear DNA Design

Long Gibson Assembly Parameters should be different than short gibson assemblies

If num fragments > 4/5 or fragment size > 8Kb, the assembly junction parameters should change (35-40bp overhangs are recommended).

Randomized tests occassionally fail

The following modules occasionally fail randomly

• Networkx Utils
• DNA complexity calculations
• Randomized designs

Bug: `Sequence is cyclic but there are no precessors` error

Describe the bug

Designs occasionally produce this bug:

dasi.exceptions.DasiSequenceDesignException: Sequence is cyclic but there are no precessors for AssemblyNode(index=0, expandable=True, type='A', overhang=False)

Unclear how to reproduce.

Add examples module users can pull from to test out DASi

Implement DNA complexity to efficiency for graphs

The efficiency of a gapped region should be lower for highly complex DNA. Highly complex DNA would be the following:

• less than 25% GC content
• higher than 75% GC content
• homopolymeric runs of 10 or more A/Ts
• homopolymeric runs of 6 or more G/Cs
• repeats greater than 14bp
• hairpins

Preferrably tests short tests in automated CI

Degenerate designs

In many cases, users may not care about the exact sequence between parts (scar sequences) or some degenerate sequence between fusion proteins (GSG linker)

DNA complexity tests sometimes fail

Better Logger

• Log to file
• Better logging levels
• Logging for CLI only

Implement sharing of primer and PCR product design

When designing multiple sequences, DASi should be able to share PCR products and primers designs in other designs.

Invalid fragment PCR products are produced.

Describe the bug
DASi can design PCR products that are not amplifiable. For example, pcr reactions that have multiple binding sites.

Expected behavior
DASi should detect bad PCRs and heavily penalize their use during the optimization process.

TODO

• make test that forces a bad PCR product.

Gap Synthesis Complexity Span Cost

PDF Assembly Reports

DASi should be able to produce nice PDF assembly reports for users.

JSON Schema display in Sphinx should be more condensed

DASi should issue a warning for highly complex repetitive sequences before transformation into bacteria

Some sequences are prone to recombination. DASi should detect this and issue a warning.

Attach notes to edges when efficiency is adjusted in Graph post processing

SBOL Output

DASi should output a SBOL file as well.

This would be a good starting point: https://pysbol2.readthedocs.io/en/latest/dbtl.html

Implement tblastn for CDS searches

In many cases, a better assembly may be found with slightly different DNA sequence that results in the same coding sequence:

• In designs, identify CDS
• use tblastn to find fragments that align to goal

Better Integration between Aquarium <> Terrarium <> DASi <> {Benchling, SynBioHub}

Improve Documentation

• Better Usage page and Installation page up front
• Better format for JSON Schema
• Point users to use output JSON format

Better explaination for DASi design failures

better parallelization

consider Design merging methods so each design can run in parallel if desired.

Clean up integration IO for DASi

Users should be able to control thresholds for valid primer pairs

• minimum bases to find primer using primer3

Find invalid assemblies

Shared sequence cost often does not match final cost

Implement DNA Synthesis Sharing

When designing multiple sequences, DASi should try to optimize the sharing of synthesized fragments to minimize cost

dasi.exceptions.DasiSequenceDesignException: Sequence is cyclic but there are no precessors for AssemblyNode(index=0, expandable=True, type='A', overhang=False)

Designs should keep record of errors (instead of raising Exception)

Annotations on SeqRecords are not being carried through to output.

Primer annotations in particular are not be carried to the final JSON output.