Core claim
portable lab
SimpleFold-3B on Apple Silicon is the proof-of-work, not the whole contribution.
First paper
preprint
Research/methods paper first; software-only paper later.
Authorship
solo
First submission defaults to Raúl as solo author and corresponding author.
Decision
The first paper should not be framed only as SimpleFold-3B runs on an M3 Max. The stronger and still reachable claim is that a single researcher can build a public, machine-verifiable reproducibility lab for local protein-structure inference. The completed SimpleFold-3B Phase 1A benchmark is the empirical anchor.
Publication Path
| Path | Decision | Reason |
|---|---|---|
| Research preprint | primary | bioRxiv/arXiv can carry new Phase 1A data and the methods argument. |
| PLOS Computational Biology / similar | target | Best fit if we emphasize reproducible methods plus biological-computational utility. |
| Bioinformatics Application Note | possible later | Needs a smoother software/service interface and broad usability. |
| JOSS/JORS/SoftwareX | companion | Good for software publication after packaging; not ideal for the first empirical-results paper. |
| Scientific Data | later dataset paper | Better when the atlas becomes a larger reusable dataset. |
Scope
Do not wait for full Phase 1B. Finish a strong SimpleFold-3B Phase 1A paper, then add limited install/runtime cards for other open systems only if they stay bounded. Those cards strengthen the thesis that open weights do not automatically imply local viability, but they should not become full target-set benchmarks before the first submission.
Authorship
The first manuscript is prepared as a solo-author submission by default. External biological or technical review is still useful, but reviewers are acknowledged with permission unless their contribution meets full authorship criteria: substantial intellectual contribution, critical manuscript revision, final approval, and accountability.
200-Hour Sprint
| Gate | Hours | Output |
|---|---|---|
| Biological and target rationale | 20 | Formal review note and target rationale table. |
| Repeated-run controls | 45 | Repeat selected hard/long targets; variability table. |
| Phase 1B-lite model cards | 40 | 2-3 install/runtime cards with precise failure or route tags. |
| Manuscript hardening | 45 | Rewrite around portable reproducibility lab plus SimpleFold-3B proof. |
| Artifact release | 30 | GitHub release, external archive, reviewer README. |
| Submission packaging | 20 | Preprint PDF, citation audit, solo-authorship and AI-assistance statements. |
Minimum Publishable Package
- 8/8 Phase 1A targets scored and public.
- Runtime card and environment JSON public.
- Biological review note complete.
- Solo-authorship policy public.
- Repeated-control limitations explicit.
- GitHub release and external archive ready.
- No future-result placeholders in the manuscript.
- Claims bounded to this model, hardware, route, and target set.
Sources Checked
- JOSS submission requirements
- Bioinformatics author guidelines
- bioRxiv screening procedures
- Scientific Data submission guidelines
- PLOS Computational Biology submission guidelines
- Benchmarking guidelines in computational biology
- Ten Simple Rules for Reproducible Computational Research
- FAIR Guiding Principles
- ICMJE role of authors and contributors
- COPE authorship-problems guidance
- Chai-1 repository
- Protenix repository
Source document: docs/35_publication_strategy_200h.md.