Alexis Stamatakis

Alexis Stamatakis

@stamatak.bsky.social

ERA Chair at Institute of Computer Science FORTH Research Group Leader Heidelberg Institute for Theoretical Studies Full Professor at Karlsruhe Institute of Technology Crete lab: https://www.biocomp.gr/ Heidelberg Lab: http://www.exelixis-lab.org/

914 followers421 following58 posts

Alexis Stamatakis
@stamatak.bsky.social

The RAxML-NG v2.0 preprint is out: www.biorxiv.org/content/10.6... v2.0 features integrated model testing, fast branch support metrics, automatic parallelization, phylogenetic difficulty prediction, genotype evolution models, to name but the most important features. And it is A LOT FASTER of course

biorxiv.orgRAxML-NG 2: Automatic model selection, novel tree search heuristics, and fast branch support metricsRAxML-NG is a widely used tool for maximum likelihood based phylogenetic inference. In the seven years since the last RAxML-NG publication, we have continuously improved and extended the code. Here, w...
Alexis Stamatakis
@stamatak.bsky.social

As computer scientist, I am very excited about our new biogeography simulator: bigrig - my PostDoc Ben Bettisworth implemented two algoritmically totally distinct simulators from scratch to conduct statistical tests for showing that bigrig is correctly implemented: www.biorxiv.org/content/10.1...

biorxiv.orgbigrig: A range simulator for the DEC[+J] modelQuality software tools for science are, as a necessity, rigorously tested and verified. However, there is a major challenge to testing software used for phylogenetics and similar analysis. There is a ...
Alexis Stamatakis
@stamatak.bsky.social

October 17 is your last chance to register for the 2nd conference on Machine Learning for Evolutionary Genomics Data (Dec 8-12), in the French Alps at legend2025.sciencesconf.org The conference talks are online at legend2025.sciencesconf.org/data/book_le...

legend2025.sciencesconf.orglegend2025 : Machine Learning for Evolutionary Genomics Data - Sciencesconf.orgEvolutionary genomics and population genetics investigate patterns of genetic diversity between species or between populations within a species and play a fundamental role in many aspects, from theoretical facets of evolution to practical ones, such as conservation genetics and biomedical sciences.
Alexis Stamatakis
@stamatak.bsky.social

New preprint: ML tree inference tools such as IQ-Tree and RAxML-NG do typically not overfit the data during tree searches: www.biorxiv.org/content/10.1...

biorxiv.orgA Systematic Investigation of Overfitting in Maximum Likelihood Phylogenetic InferenceMaximum Likelihood (ML) tree inference reconstructs phylogenies from Multiple Sequence Alignments (MSAs). Since MSAs are inherently noisy, ML tools may experience overfitting, whereby the inferred top...
Alexis Stamatakis
@stamatak.bsky.social

Check out our new preprint on reproducible parallel phylogenetic inference under varying core counts - it also includes a generic method for reproducible parallel associative reduction operations www.biorxiv.org/content/10.1...

biorxiv.orgBit-Reproducible Phylogenetic Tree Inference under Varying Core-Counts via Reproducible Parallel Reduction OperatorsMotivation: Phylogenetic trees describe the evolutionary history among biological species based on their genomic data. Maximum Likelihood (ML) based phylogenetic inference tools search for the tree and evolutionary model that best explain the observed genomic data. Given the independence of likelihood score calculations between different genomic sites, parallel computation is commonly deployed. This is followed by a parallel summation over the per-site scores to obtain the overall likelihood score of the tree. However, basic arithmetic operations on IEEE 754 floating-point numbers, such as addition and multiplication, inherently introduce rounding errors. Consequently, the order by which floating-point operations are executed affects the exact resulting likelihood value since these operations are not associative. Moreover, parallel reduction algorithms in numerical codes re-associate operations as a function of the core count and cluster network topology, inducing different round-off errors. These low-level deviations can cause heuristic searches to diverge and induce high-level result discrepancies (e.g., yield topologically distinct phylogenies). This effect has also been observed in multiple scientific fields, beyond phylogenetics. Results: We observe that varying the degree of parallelism results in diverging phylogenetic tree searches (high level results) for over 31 % out of 10 130 empirical datasets. More importantly, 8 % of these diverging datasets yield trees that are statistically significantly worse than the best known ML tree for the dataset (AU-test, p < 0.05). To alleviate this, we develop a variant of the widely used phylogenetic inference tool RAxML-NG, which does yield bit-reproducible results under varying core-counts, with a slowdown of only 0 to 12.7 % (median 0.8 %) on up to 768 cores. We further introduce the ReproRed reduction algorithm, which yields bit-identical results under varying core-counts, by maintaining a fixed operation order that is independent of the communication pattern. ReproRed is thus applicable to all associative reduction operations – in contrast to competitors, which are confined to summation. Our ReproRed reduction algorithm only exchanges the theoretical minimum number of messages, overlaps communication with computation, and utilizes fast base-cases for local reductions. ReproRed is able to all-reduce (via a subsequent broadcast) 4.1 · 106 operands across 48 to 768 cores in 19.7 to 48.61 µs, thereby exhibiting a slowdown of 13 to 93 % over a non-reproducible all-reduce algorithm. ReproRed outperforms the state-of-the-art reproducible all-reduction algorithm ReproBLAS (offers summation only) beyond 10 000 elements per core. In summary, we re-assess non-reproducibility in parallel phylogenetic inference, present the first bit-reproducible parallel phylogenetic inference tool, as well as introduce a general algorithm and open-source code for conducting reproducible associative parallel reduction operations. ### Competing Interest Statement The authors have declared no competing interest. European Research Council, https://ror.org/0472cxd90, 882500 European Union, https://ror.org/019w4f821, 101087081
Alexis Stamatakis
@stamatak.bsky.social

RAxML-NG v2.0 beta is here github.com/amkozlov/rax... Features 1. Difficulty prediction academic.oup.com/mbe/article/... 2. Adaptive search academic.oup.com/mbe/article/... 3. Early Stop www.biorxiv.org/content/10.1... 4. Model selection 5. Rapid bootstrapping academic.oup.com/mbe/article/...

github.comRelease RAxML-NG v2.0-beta2 · amkozlov/raxml-ngMain new features Adaptive tree search heuristic (Togkousidis 2023) , (Togkousidis 2024) Multiple fast branch support measures Integrated model testing Full Changelog: 1.2.2...2.0-beta2
Alexis Stamatakis
@stamatak.bsky.social

Check out raxtax, our new open-source tool for taxonomic classification of barcoding sequences, it's 2.7-1000 times faster than competing tools and also implements fancy uncertainty scores: www.biorxiv.org/content/10.1...

biorxiv.orgraxtax: A k-mer-based non-Bayesian Taxonomic ClassifierTaxonomic classification in biodiversity studies is the process of assigning the anonymous sequences of a marker gene (barcode) to a specific lineage using a reference database that contains named seq...
Alexis Stamatakis
@stamatak.bsky.social

Are you analyzing genotype data via PCA or MDS, in particular including ancient DNA samples? Here is a novel easy-to-use tool to assess the stability of these analyses by bootstrapping the SNPs: academic.oup.com/bioinformati...

academic.oup.comPandora: A Tool to Estimate Dimensionality Reduction Stability of Genotype DataAbstractMotivation. Genotype datasets typically contain a large number of single nucleotide polymorphisms for a comparatively small number of individuals.
Alexis Stamatakis
@stamatak.bsky.social

News article about the phenomenon of specimen-drain from the global South to the global North by award-winning science journalist and good friend Vasiliki Michopoulou - on 3 out of 35 ancient DNA papers with Greek samples, Greek scientists were first and/or last authors www.dnews.gr/eidhseis/sci...

dnews.grΜετά το brain - drain η Ελλάδα βιώνει και τη διαρροή δειγμάτων για επιστημονική έρευνα - DnewsΠαρότι αρκετές ελληνικές επιστημονικές ομάδες θα μπορούσαν να ηγηθούν ερευνητικών έργων λόγω του πλούτου των αρχαιολογικών και παλαιοντολογικών δειγμάτων της χώρας, συχνά περιορίζονται στον ρόλο του σ...
Alexis Stamatakis
@stamatak.bsky.social

Another new term we invented together with a colleague is specimen-drain: the export of valuable biodiversity, ancient DNA, or other samples to countries of the global North that have money to process them coupled with losing the lead on the respective research papers

Alexis Stamatakis
@stamatak.bsky.social

Last reminder - applications for our 2025 school on computational molecular evolution in Crete close in one week - there will be no application deadline extensions. meetings.embo.org/event/25-com...

meetings.embo.orgComputational Molecular EvolutionThe need for effective and informed analysis of biological data is increasing with the explosive growth of genomic data. A phylogenetic framework is central to many molecular evolutionary approaches …
Alexis Stamatakis
@stamatak.bsky.social

Applications are still open for our 2025 school on computational molecular evolution in Crete, the application deadline is November 1st - only about 10 days from now. meetings.embo.org/event/25-com...

meetings.embo.orgComputational Molecular EvolutionThe need for effective and informed analysis of biological data is increasing with the explosive growth of genomic data. A phylogenetic framework is central to many molecular evolutionary approaches …
Alexis Stamatakis
@stamatak.bsky.social

Before somebody else comes up with it, here is a new term I invented: brain-redrain Definition: brains that return to their underdeveloped home country in the hope to help improving things but then leave again since they are frustrated because nothing will ever change there.

Alexis Stamatakis
@stamatak.bsky.social

Do you want to eliminate subjective choices (and a lot of manual work) in cognate data synonym selection when building language phylogenies? Check our new approaches for avoiding synonym selection: arxiv.org/abs/2404.19328

arxiv.orgComputational Approaches for Integrating out Subjectivity in...Working with cognate data involves handling synonyms, that is, multiple words that describe the same concept in a language. In the early days of language phylogenetics it was recommended to select...
Alexis Stamatakis
@stamatak.bsky.social

Do you want to reduce the CO2 footprint of your computations and save money at the same time via adaptive CPU frequency scaling? Check out our new open-source EcoFreq tool: ieeexplore.ieee.org/document/105...

ieeexplore.ieee.orgEcoFreq: Compute with Cheaper, Cleaner Energy via Carbon-Aware Power ScalingHigh-performance computing (HPC) constitutes an energy-hungry endeavor, and any efficiency gains via hardware and software advances are quickly (over-)compensated by increased consumption (rebound eff...
Alexis Stamatakis
@stamatak.bsky.social

Recent collaborative forensic work with the aDNA lab to identify the victims of a WW2 mass execution in the village of Adele, Crete: www.sciencedirect.com/science/arti...

sciencedirect.comIdentification of the 18 World War II executed citizens of Adele, Rethymnon, Crete using an ancient DNA approach and low coverage genomesIn the Battle of Crete during the World War II occupation of Greece, the German forces faced substantial civilian resistance. To retribute the numerou…