Conservative essentiality evidence classes

HANSEN Essentiality Prediction

Ranked Mycobacterium leprae proteins by conservative essentiality evidence, integrating ProteomeLM contextual signal, ESM-2 transfer-learning probability, Mtb ortholog anchor evidence, HANSEN target-priority evidence, functional annotation, pathway context and confidence-based review flags.

Conservative interpretation

Proteins are grouped into evidence classes designed to avoid over-interpreting low-evidence predictions.

Proteins

—

Likely Essential

—

Possible / Review

—

Uncertain

—

Lower Evidence

—

Per-protein Conservative Essentiality Evidence

The table uses conservative evidence classes rather than hard non-essential calls. Use “Possible essential / review” for manual inspection.

Download TSV

Rank ?	ML ID ?	Gene ?	Protein ?	Evidence Class ?	HANSEN Score ?	ESM-2 P ?	Mtb Anchor ?	ProteomeLM ?	Target Priority ?	Confidence ?	Rationale ?
Loading…

How this essentiality evidence is calculated

HANSEN combines an ESM-2 Mtb-transfer classifier trained using DeJesus et al. (PMID: 28096490) calls, reciprocal-best-hit Mtb anchors, ProteomeLM context and HANSEN target-priority evidence.

Read methods in Help

1. About HANSEN

HANSEN is a research platform for the Mycobacterium leprae proteome. It aggregates protein structure predictions, binding-pocket predictions, model-quality metrics, epitope predictions, target prioritisation and gene-essentiality predictions. The platform integrates independent third-party scientific tools, each governed by its own upstream licence.

By accessing or using HANSEN, users agree to these Terms and to the upstream terms of every tool whose outputs they access, generate, view, download or redistribute through HANSEN. These Terms do not override upstream licences. Where a stricter upstream term applies, the stricter term governs.

2. Permitted Use

HANSEN is provided for non-commercial scientific research and educational use only. This platform-wide restriction is driven by the most restrictive integrated component, currently AlphaFold 3 where its assets or outputs are present.

Permitted users include universities, non-profit research institutes, educational bodies, and public-health or government bodies acting in a non-commercial capacity.
Commercial use, contract research for a company, or use by or on behalf of a commercial organisation is not permitted where AlphaFold 3 assets or outputs are involved unless separate permission or licensing has been obtained.

3. Research Use Only — No Clinical or Diagnostic Use

HANSEN outputs are computational predictions for hypothesis generation and research prioritisation only. They are not experimental findings and must not be used to make, support or inform clinical, diagnostic, therapeutic, legal, employment, insurance or other decisions affecting any individual.

Confidence metrics such as pLDDT, PAE, model confidence and ranking scores describe model self-assessment. They are not guarantees of biological, diagnostic or therapeutic validity.

4. Integrated Third-Party Tools and Licence Links

Attribution and citation are mandatory for academic outputs. Users must cite the relevant upstream tools when HANSEN-derived results are published, shared or redistributed.

Area	Tool / Resource	Link	Use Notes
Structure modelling	AlphaFold 3	Repository Output Terms Model Parameters Terms	Non-commercial use only; AlphaFold 3 parameters must not be redistributed through HANSEN.
Structure modelling	Boltz-1 / Boltz-2	Boltz Repository	MIT-licensed code and weights according to the upstream repository; cite Boltz reports when used.
Structure modelling	Chai-1	Chai-lab Repository	Apache 2.0 according to the upstream repository; cite Chai-1 technical report when used.
Binding-pocket prediction	AF2Bind / AlphaFold 2-derived workflows	ColabDesign AlphaFold DB	Verify the exact AF2Bind repository and AlphaFold 2 / ColabDesign terms before redistribution or commercial use.
Binding-pocket prediction	P2Rank	P2Rank Repository	MIT-licensed according to the upstream repository; cite Krivák and Hoksza.
Binding-pocket prediction	fpocket	fpocket Repository	MIT-licensed according to the upstream repository; cite Le Guilloux, Schmidtke and Tufféry.
Epitope mapping	DiscoTope-3.0	DTU DiscoTope-3.0 Server Academic Download / Licence	Academic / non-commercial use; commercial use requires a separate DTU licence.
Target prioritisation	ProteomeLM / ProteomeLM-Ess	ProteomeLM Repository	Check ProteomeLM and ESM-C terms before redistribution or commercial use.
Gene essentiality	ESM-2	ESM Repository	MIT-licensed according to Meta AI ESM repository; cite Lin et al.
Gene essentiality	DIAMOND	DIAMOND Repository	GPL-3.0. Keep as an external dependency unless GPL obligations are fully addressed.
Machine learning	scikit-learn	scikit-learn Citation	BSD-3-Clause; cite scikit-learn where used.

5. AlphaFold 3-Specific Obligations

AlphaFold 3 assets and outputs are for non-commercial use only.
AlphaFold 3 model parameters must be obtained directly from Google by an authorised user and must not be redistributed through HANSEN.
Any AlphaFold 3 output displayed, exported or distributed through HANSEN must retain the required legally binding terms-of-use notice.
AlphaFold 3 outputs must not be used to train or create machine-learning models for biomolecular structure prediction, including by distillation.
AlphaFold 3 outputs must not be used for healthcare, diagnostic or individual-rights decisions.

6. Output Ownership, Attribution and Redistribution

Tool providers generally do not claim ownership of original outputs generated by users, but several reserve rights and may allow identical outputs to be generated for other users. Users obtain no exclusivity in any prediction.

Any publication, presentation or onward distribution of HANSEN-derived results must cite every tool involved and must preserve any upstream notices required by the relevant licences.

7. Acceptable Use

Users must not:

misrepresent the origin of integrated tools or outputs;
imply endorsement by upstream tool providers;
circumvent access restrictions on model parameters;
upload content they are not authorised to process;
use HANSEN for unlawful purposes or in breach of upstream licences or prohibited-use policies.

8. Privacy Notice

HANSEN may process scientific data including protein sequences, structures, model files and prediction parameters. These are not personal data unless linked to an identifiable individual. Users should not upload human-identifiable data unless separate ethics, information-governance and lawful-basis arrangements are in place.

If accounts or logs are enabled, HANSEN may process user names, institutional emails, organisation names, authentication metadata, IP addresses, timestamps, accessed endpoints and queries. These may constitute personal data under UK GDPR and the Data Protection Act 2018.

Depending on deployment, some modelling workflows may transmit uploaded sequences or structures to external third-party services, for example public MSA servers or public prediction servers. For unpublished or sensitive data, integrated tools should be run locally or on University-controlled infrastructure wherever possible.

HANSEN does not sell personal data and does not use it for advertising.

9. Storage, Security and Retention

Research inputs and generated outputs are retained according to the research project, funder and University requirements. Access should be restricted to authorised research-team members, with reasonable technical and organisational controls to protect against unauthorised access, loss or disclosure.

10. User Rights and Contact

Where HANSEN processes personal data, individuals may have rights under UK data protection law, including rights of access, rectification, erasure, restriction, objection and data portability, subject to applicable exemptions for scientific research.

11. Changes

These Terms may be updated as HANSEN changes or as upstream tool licences change. Users should re-check the current terms before redistributing HANSEN-derived outputs or using HANSEN in any external collaboration.