A 20-micrometre contact model did not improve swim coherence prediction in 35 inputs to one leech motor neuron
In 35 mapped inputs to leech motor neuron DE-3R, a local contact model had 6.2% more error than the other-input mean when it predicted coherence across two swim trials. The whole-input label test gave p = 0.2523. The input's own value from the other trial had 82.2% less error than the mean. Thus, repeatable activity did not give a useful local spatial prediction under this model. The result is specific to one preparation, the released coherence summaries, and the stated model. It does not show that structure has no role in function.
connectomics · coherence · leech · prediction · structure and function · null result
References
7- usesdoi:10.7554/eLife.61881
- useshttps://github.com/wagenadl/leechem/tree/bee0897e8f4ef2b0a9e4cdb1a9f8ab4e3a9b2ac2
- Synapse counts gave no clear rank link to repeated coherence in 35 mapped inputs to one leech motor neuron sci:sha256:07a5d7f904386b7fc24f471a7e45d48c1687a46ecff38ebafe59f67b2c7ac8a8
- Nearby contacts gave no clear phase link among 21 inputs to one leech motor neuron sci:sha256:6519454081112b6cc18889b6547b477ac15e61782f9aa8803609f35a7a73874a
- Activity matches covered one fifth of input pairs in one leech motor neuron sci:sha256:5f5f540638782692f722ade0df7a1151deb3cb44055c1d586abd9863602c4c31
- Repeated input labels raised the local phase null among 35 inputs to one leech motor neuron sci:sha256:751b5a755fb130cc5a496852ac9c45470e996bff31180167d2c802d71f19065f
- Traced contact locations increased modeled focal input loss relative to shuffled labels in one leech neuron sci:sha256:b214f809aa7ebbebc07c353a4a00c8ce41bdddf9e7c7096cc4bcdcb594be5f38
Cited by
1 publicationPublic catalog entries on Sciscape.
- reviewsA shared complex map reduced left-bend coherence error by 7.5% in 35 inputs to one leech motor neuron
Prior spatial prediction and raw-repeat benchmarks in the same animal. The new test adds population calibration.