iiinova — where your digital self keeps evolving

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 01/10/2026

I like it better when my mountains are not top international news 😢

Front page of The Guardian app, with an image of a glacier landscape with melting ice, highlighting a large crevice with deep blue water visible. The news headline is
‘Explosion of melt rates’: 20% of Swiss glaciers lost in five years

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 04/09/2026

From this perspective on AI and "peer" review, I'm surprised to actually not see any increase in the rate of growth of articles with LLMs? www.science.org/content/article/can… #AcademicPublishing #PeerReview #AI

Bar graph of the number of scientific publications per year since 1990. It shows a continuous increase. The introduction of ChatGPT in 2022 is highlighted.

112 2

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 29/07/2026

New preprints! 1/3 Spider venom glands are modified salivary glands which have recruited appendage and myogenic programs from other the developmental programs of other organs. Led by Giulia Zancolli, with @yehumoran #EvoDevo #MolecularEvolution […] [Original post on ecoevo.social]

Schematic view of the embryonic expression pattern of the genes Sage, Sum1 and Dll in a spider embryo, showing Sage only in venom gland primordia, Sum1 in chelicerae and other appendages, and Dll at the distal end of appendages

100 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 10/07/2026

Real colors satellite photo shows the state of vegetation after weeks of heat wave and drought #ClimateChange #drought #heatwave

Satellite photo of France and surrounding Europe showing yellow brown vegetation except on mountains.

011 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 09/07/2026

There are often legitimate complaints about the fun photos illustrating news articles on climate change and heat waves, so here's @TheGuardian doing it well.

Title "Climate crisis
Western Europe records hottest-ever June as heatwaves intensify"
Illustrated by photo of miserable looking cows among dried vegetation.

072 7

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 02/07/2026

After @GaltierNicolas 's #SMBE2026 talk I of course calculated my @DAFNEE score. It is notably brought down by NAR Database 🤷

histogram of Dafnee scores showing my score at 65%

041 4

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 29/06/2026

The poor wifi also limits the usefulness of the #SMBE2026 app, which always wants to reload and often doesn't manage. 😕

100 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 29/06/2026

Lin #SMBE2026: a high noise mutant enhances cell competitiveness under a fluctuating environment. Depending on backgrounds, noise is advantageous or deleterious.

Speaker showing a slide with two side-by-side models.
Left, a high threshold is lethal, noise is bad.
Right, a low threshold is lethal, noise is good.

000 0

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 09/06/2026

Welcome to Hogwarts for the main #EED2026 meeting. Already got a bit lost among the moving staircases during the satellites.

Ceiling of gothic hall at university of Glasgow

010 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 30/05/2026

After 2 days with school classes, ready to explain moulting to families at University of Lausanne over the weekend, from fossils to citizen science.

Foreground tablets open on iNaturalist and moulting.org
Background tables with fossils and moults

121 2

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 02/04/2026

Are you going to #EED2026 Glasgow by train, or are you hesitating? Let's share experiences and give ideas to those who think to fly from Western Europe with the hashtag #NoFlyEED. With the Eurostar, if you can get to a station it serves, then it's easy to get from London to Glasgow. ✅🚆 ❌✈️

List of populat stations served by Eurostar to London: Paris, Brussels, Amsterdam, Lille, Rotterdam, Disneyland, Cologne, Schiphol

032 3

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 17/03/2026

I've discovered that there are 8 rodent species for which I deposited the 1st ever DNA data in NCBI. Each nucleotide was handcrafted with manual sequencing and acrylamide gels. www.ncbi.nlm.nih.gov/sites/myncbi/m…

000 0

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 06/03/2026

Here this sum over organs/tissues shows a strong excess of positive selection on regulatory sequences in human nervous system evolution 😎. And to a lesser extent in male reproductive system. #MolecularEvolution #EvoDevo

Graph showing the SUMSTAT statistic for different organs or tissues, colored by system. Values from the Nervous and Male reproductive systems have the highest values, while all others are quite similar.

030 3

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 02/03/2026

@nextstrain The new observations, coupled with collaborations with the teams of Allison Daley and Ariel Chipman, show trends of convergence among species similar head shields.

Phylogenetic distribution of egress direction during moulting and ecdysial suture position across major arthropod lineages.
Left, simple phylogeny of athropod groups, with red dots indicating occurrence of taxa characterised by a domed and pronounced horseshoe-shaped head shield in Radiodonta, Trilobita, Xiphosurida, and Brachiopoda.
Right, table of occurence of traits of egress direction and of ecdysial suture position.

010 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 22/01/2026

New horseshoe crab reliefs have joined the isopods on my office wall 🤩

Photo of a wall with a photo of an isopod in 3 positions on top, and 3 reliefs of a horseshoe crab below

032 3

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 15/12/2025

Beaucoup de grippe en ce moment en Suisse, et ça n'est pas fini ! Vaccinez-vous si possible, et soyez prudents. #GrippeSuisse #grippe

Overlaid graphs of influenza A incidence in Switzerland, with previous years in grey, and this year so far in purple, aligned per date. This years' graph is the highest so far for December.

000 0

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 03/12/2025

Interestingly, the peaks under positive selection are enriched in embryo and adult, and the associated genes are more expressed in reproductive and immune-associated organs.

barplot of proportion of directional selection for peaks at different developmental stages, showing ≈6% for embryonic and adult, 5% for larva, and less than 2% for prepupa and pupa.

Plot of fold enrichment for different anatomical terms, showing high enrichment (in decreasing order of FDR significance) for testis, fat body, digestive system element, Malpighian tubule, seminal fluid secreting gland, midgut and ovary.

100 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 03/12/2025

As expected from pop gen, the peaks inferred under positive selection have lower polymorphism and higher divergence (≈ MK test). The associated genes have a very weak association to negative selection on the coding sequence.

6 pannels A-F showing results on real Drosophila melanogaster data.
A: violin plots of proportion of peaks inferred under directional selection, with a median around 20% for the permutation test and 5% for RegEvol.
B: positive relation between ΔSVM quantile and proportion of peaks inferred under directional selection.
C: bar plot of ratio of substitutions to polymorphisms for one experiment, showing a higher ratio for peaks inferred under directional selection than for those not.
D: volcano plot of -log10(pvalue) as a function of odds ratio showing that peaks inferred under directional selection almost all have a higher ratio of substitutions to polymorphisms.
E: negative relation (rho=-0.22) between the number of peaks associated to a gene and its omega0.
F: weak positive relation (rho=0.08) between the proportion of peaks associated to a gene inferred under directional selection and its omega0.

100 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 03/12/2025

Importantly, RegEvol is robust to ascertainment bias (see also sup fig 8), rare extreme effect substitutions, and high sequence divergence, unlike the permutation test.

$False positive rate as a function of: ΔSVM under neutrality or under stabilising selection; number of substitutions under stabilising selection; fraction of dataset (highest SVM). In all cases RegSel has almost 0 false positives, while false positives increase for the permutation test with more extreme data.$

100 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 03/12/2025

Simulations show that RegEvol detects both selection to increase or decrease (but with less power) transcription factor binding, better than the previous permutation test.

pannel A: barplots of proportion of inferred models under different simulated scenarios, showing that RegSel and the permutation test mostly infer the correct model, but with more false negatives in the permutation test, which also infers neutral evolution for simulation under stabilizing selection.
pannels B-E: true positives of both tests as a function of: substitution number, ΔSVM, proportion of neutral substitutions, strength of directional selection. Power increases with each except with neutral substitutions where it decreases. In all RegSel has more power.

100 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 03/12/2025

The model thus uses the distribution of ΔSVM of each peak, from all possible substitutions (instead of a random subset as in the previous method), which is multiplied by a selection regime to obtain an expected distribution of fitness effects per peak. The fit […] [Original post on ecoevo.social]

Graphical summary of the evolutionary model, in 4 rows, A to D.
A: left, bell shaped distribution of phenotypic changes; right, observed substitutions between an ancestral sequence and a modern one, using an outgroup.
B: 4 fitness landscapes: no selection, alpha=beta=1, uniform distribution; stabilizing selection, alpha=beta≠1, bell shape; directional selection, either alpha>beta, curve skewed right, or alpha<beta, curve skewed left.
C: 4 distributions which are the product of the distribution in A and each of the fitness landscapes. Observed subtitutions are mapped onto them, falling outside of the high density area except for the third, i.e. directional selection alpah>beta.
D: cartoon likelihood indicating that only the third fitness landscape is supported.

110 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 03/12/2025

We first show that the ΔSVM score captures functional signal of transcription factor binding and its evolution. OTOH conservation-based scores such as phastCons or phyloP do not correlate with function. ΔSVM thus provides a good basis to estimate of distributions of fitness effects.

Three pannels, A to C.
A: 7 rows: DNA sequence with letters proportional to SVM; all DNA changes with letters proportional to ΔSVM; coverage; SVM then ΔSVM showing that they are both highest at the maximum coverage; phastCons then phyloP showing that their variation is unrelated to coverage.
B: 5 rows, each showing the mean distribution of parameters over many peaks: coverage peaks in the center (by definition), SVM and ΔSVM have sharp peaks in the center, phastCons has a very broad peak in the center, phyloP has no peak.
C: violin plots of distribution of Pearson correlations with peak coverage: SVM and ΔSVM have mostly positive correlations, phastCons and PhyloP have correlations centered on 0.

100 1

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 03/12/2025

I'm excited to present this new model to detect positive selection on regulatory sequences, which has been 3 years in the making! Thanks to Alexandre Laverré and @Phylogenetrips for their amazing work on this project. 😃 doi.org/10.1101/2025.11.26.690685 #bioinformatics #molecularEvolution

33618 36

Marc Robinson-Rechavi @marcrr.ecoevo.social.ap.brid.gy · 08/09/2025

Great to see the Mastodon and Bluesky but *not* X/Twitter handles on the intro slide of @dessimoz @SIB #bc2basel

133 3