临时改签机票，提前一天回到北京。这样，本来计划在火车和自然历史博物馆完成的剑桥会议记录没有能够及时完成。今天终于有时间初步整理一下，就以当时电脑记录的文档和部分照片为主，和大家分享一下参会的体会。 

本次得到中国科学院国际合作局的批准访问英国，我、Michael Orr博士和曹焕喜同学到英国学术访问。剑桥大学是继英国自然历史博物馆（Michael Orr和曹焕喜）、北安普顿大学（Michael Orr和我）、伯明翰大学（我本人）之后的最后一站。剑桥大学动物学系和遗传学系举办的会议是此次学术访问的重点。Michael Orr博士在访问Jeff Ollerton教授和Paul William博士之后，决定参加和他专业方向更加接近的熊蜂会议。曹焕喜则在检视完自然博物馆部分姬小蜂标本后，18日中午到King's Cross和我会合，直奔剑桥大学。到了剑桥，我们得到老朋友谭声江博士的热情接待，并参观了部分研究单位。声江原来在动物研究所获得博士学位，后到剑桥和牛津做博士后，现在定居在剑桥。

曹焕喜在国内通过网络预订了Fairways Guest House。该店在网上评价比较好，离火车站走路大约10分钟，离市区大约30分钟。我住的客房在阁楼，简约卫生而清爽，还配装了数字电视。房客比较多，但都比较安静。

3月19日一早，我们就从入住的家庭旅社Fairways Guest House前往St. John's College。初春天气，加上刚刚下过雪，剑桥天气还是比较冷。由于路线不熟，Google地图把我们引导到了不同的College，还是很费了一点时间。不过也好，我们可以沿路欣赏剑桥的街道和美景。后来我们才发现，会程安排紧凑，内容精彩；咖啡和午间冷餐时间都很短；晚上则在餐厅安排了晚饭。此外根本没有闲暇走入大学或者街道。早晨沿路赏景是唯一的选项。

Speciation Genomics Symposium期间，我见到了Roger Butlin。我2004年3月访问Leeds大学时，曾经去他办公室简短交谈。后来，他很快就带领整个实验室搬到了Sheffield大学，带走了一个学科。后来在英国皇家昆虫学会年会，我第二次在South Kensington见到他。此次会议，显然他是幕后的大家之一；而相对年轻的科学家则走到前台。蝴蝶斑块的基因决定，鱼类基因组在物种分化研究中的应用是此次会议报告亮点。同时，从会议报告的安排来看，几乎所有的报告内容都涉及到物种分化的模型和数据模拟。这些工作非常符合基础研究的特点：寻找并试图解决根本的生物学问题。部分工作，我们已经通过阅读文献，初步掌握了部分方法和内容。但这些报告人及其所在团队的总体思路和核心技能，我个人认为还有待进一步了解和学习（笔记附后）有一个报告，是我最希望了解的内容之一。可惜报告人的英语实在难懂，我只能抓住零星几个单词。只能计划在后续他的论文中继续理解其精华了。由此，有好的思路和工作成果，还需要包括语言在内的较好交流能力，我们才能把我们的工作做得更好。

我参加过中美、中德和我们自己组织的研讨会panel discussion。每次主办方都会围绕主题，提出一些学科方向的重要问题，邀请参会方或者知名科学家进行讨论。此次会议方准备了很多问题，但看起来效果不是特别好，讨论的氛围也不是特别热烈。总体而言，Roger等台上的科学家讨论得比较多，而台下听众显然很投入，但互动不是特别活跃。

举办方的想法主要立足剑桥大学，面向欧洲，邀请美国等其他地区的学者。会议规模有所控制。我们是唯一来自亚洲的参会人员。此次会议我们没有准备学术报告，带着学习和熟悉会议规则的态度积极前往参与。显然，国内很多学者的工作已经达到或者超过了部分报告人的水平。整体而言，这个团体在多年的积累下，已经自然形成了更高的水平。但是，目前该群体也遇到了问题。其中最大的挑战就是科研经费的削减：除了癌症和干细胞研究相对可以获得较高的资助外，进化生物学领域获得经费较少，资助率也很低。

图1、家庭旅社Fairways Guest House

图2、路上的博物馆之一

图3、河边初春

图4、终于接近会议地点

图5、主持人登场

图6、咖啡时间

图7、Panel Discussion

图8、看来问题很多、很难

Mark

Species concepts and the modern synthesis

Hybrid zones and gene flow, Barton & Hewitt (1989), Nature

:balance between dispersal and selection against hybrids

Genic concept of speciation

:divergent loci resist gene flow, Wu (2001) JEB; Wu & Ting (2001) Nat. Rev. Gen

Gene flow continues but linkage builds and divergent regions grow; complete reproductive isolation evoles

The rise of speciation islands, Turner et al. (2005) PLoS Biology; Turner & Hahn (2007) MBE

:Anopheles gambiae, M+S forms

Speciation islands and processes, Nosil et al. (2009) Mol Ecol; Feder et al. (2012) Trends in Genetics

Continuums and islands, Martin et al., (2003) Genome Research; Seehausen et al., (2014) Nat. Rev. Gen.

Mirages and alternative explanations, Noor & Bennet (2008) Heredity; Cruickshank & Hahn (2014) Mol. Ecol.

:background selection;

:local adaptation after isolation;

:shared ancestral polymorphism

Linked selection and recombination, Burri et al., (2015) Genome Res.

Confounding factors, Fst (picture required)

The stickleback speciation continuum

Where do we go from here?

From speciation genes to processes building reproductive isolation; more confounding factors into models; shifting more from description

Reto:

The dynamic buildup of heterogeneous differentiation landscapes by linked selection

A critique of the adaptationist program

Avoiding molecular ‘spandrels’, Ravinet et la., 2017

LS: purifying (background) and positive selection

Heterogeneous genome-wide LS

Impact LS on diversity, Begun & Aquadro (1992);

Relative measures of differentiation: Charlesworth 1998, Cruiskhank and Hahn 2014

Speciation islands, mirage in the desert: Ellegren et al., 2012

: reduced diversity

:consistent with LS

:conserved LS -> correlated genomic landscapes

The system: bush monkeyflower radiation

:chromosome-level genome; genome re-sequencing

:very similar levels of diversity within all taxa

PC1 Fst PC1 pei mirrow one another

Levels of diversity are correlated with gene diversity and recomb. Rate

The heterogeneous signature of LS should build gradually and over time (simple model of allopatric divergence)

Conclusions – across a plant system; differentiation landscape; heterogeneous

Anja

Understanding the genomic basis of speciation with gene flow using hybrid zone analysis

Identify regions affected by divergent selection & genomic locations

:hybrid zone; cline analysis; geographical ‘replicates’

:allele frequency differences; cline width; cline centre;

Rocky shores – Littorina saxatilis: selection pressures – wave action; crab predation

Sampling sites – ANG (Sweden)

~138000 biallelic SNPs, 50% show significant clines

‘var.ex’ refelecting the cline width and Fst

Simulations -. Neutral distribution of var.ex

Outliers within linkage groups

Conclusions – balancing selection/inversion; deviate from simple model of divergent selection (not detectable by Fst outlier scans, but cline analysis)

Geographical replicates

Contact zones; islands < 10 km apart

Expectation: less sharing expected with Fst scans than with cline analysis

More contact zones – hierarchical sampling design

Only work on top 1% SNPs (664): large proportion of cline outliers shared

More false positives among Fst outliers

Conclutions – 70% shared outliers are in inversions

Joana

Sources of genetic variations

:new mutation; standing variation; hybridization

Ancient hybridization

Different species, different genomic mosaic of the parental lineages

:red opsin gene, adaptation to different water depth, Meier et al. 2017, Nat. Comm.

150,000 years ago: hybrid origin

19000-15000 yrs ago: lake Victoria completely dry

Did the 500 endemic cichlid species really evolve from a single species in 15000 yrs?

Whole genomes resequencing > 450 genomes -> PCA of 152 genomes

Mitochondrial genome tree reveals two deeply divergent lineages in lake Victoria

Lake Victoria species mainly cluster by Genus/ecology

(2.7 M SNPs, 293 genomes) PC1 (6.15%)

Pundamilia species complex, Meier et al. 2017, Mol. Ecol.

:parallel differences in phenotype, water depth and color vision

(Seehausen et al., 2008; Seehausen, 2009)

Demographic modeling of the whole genomes reveal speciation from hybridization

Recent hybrid parallel speciation

Many regions are highly differentiated between the species, Meier et al., accepted at MBE

Fst between species pairs are very different but high differentiation regions are

Enrichment of selection statistics (various ones) support the action of selection

TWISST (Martin & sb, 2017)

Sorting of admixture variation under parallel selection pressures

Highly differentiated regions are associated with low recomb. rates

Very old haplotypes segregate in Lake Victoria cichilids at shared highly differentiated regions

Simon Martin

Predicability/genomic architecture/species barriers

Strength of the species barriers: Barto & Bengtson 1986; Mallet 1995; Wu 2001 et al.

Architecture of the barrier: genes? Few or many loci; distribution in genomes; small or large effects

Characterising the species barrier: the dream

Helicornius “races”

Predation pressures on races: optix; wnt-A; cortex

Islands of divergence, Martin et al., 2013, Genome Res.

Are islands predicatable? Van Belleghem et al., Nature Ecol. Evol.

Repeted divergence of regulatory modules of the same genes

What about ‘good’ species?

Habitat preferences; host plant choice; hybrid females sterility (Haldane’s Rule); disruptive selection

Parapatric races; sympatric species, Martin et al., 2013; Seehausen et al., 2014

What do we want to measure?

:effective migration rate/admixture proportion

Fd: estimated admixture proportion: sympatric/allopatric, Martin et al., 2015 MBE

Species barriers in two separate locations: strong as predicted on the Z

Are these polygenic species barriers: What do we expect to see ( foreign chromosome – recomb., selection)?

Recomb. predicts barrier strength: Borton & Bengsson 1986; Aechbacher et al., 2017

Recomb. predicts admixture: polygenic species barriers

Are the loci underlying polygenic barriers predictable?

Stable recom. Rates make the question intractable: Davey et al., 2017, Ecology Letters

Recomb. also varies at the chromosome scale: longer chrom., stronger barrier

Reduced recomb. at chromosome ends: Barton & Bengtsson 1986

Unpredicatable barriers near chromosome ends?

Is the role of structural variation in speciation predictable? Davey et al., 2017, Evolution Letters – No inversions (nor any recomb. suppression)

Predictability across genera? Danaus chrysippus species complexes; localized blocky species barriers; putative inversions included ~ 800 genes; a different sets of wing patterning genes; repeated insertion of regions; deletions of regions

DISCUSSIONS –

How much do you know about the pigmentation development/pathway?

How many genes have been involved in the pigmentation?

Response to selection pressures?

Camille

Molecular divergence and genetic isolation

Speciation: evolution of reproductive isolation

Causes of postzygotic isolation – decreased recomb. with divergence d; accumulation of Dob-Mul. Incompatibilities with d; hybrids can be unadapted to parental environments

From one to two species – when can we detect the effects of the first barriers on gene flow? Is there a threshold of divergence above which gene flow is impossible?

Questions – comparing alternative scenarios for 61 pairs of species along a continuum of d (ABC)

Testing for introgression between species using ABC

Explored range of divergence

General relation between d and ongoing migration

Explored range of divergence in Helioconius

Results of model comparisons over 28 pairs – 3 pairs of populations over 4; 2 pairs of sympatric over 4; 0 pair of allopatric ove 20

Any species barriers?

:locus specific model comparison; how 303 loci with a reduced me are distributed? Producing 9 bins of 1,170 genes; genomic distribution of the 303 loci with a reduced me;

Doro Lindtke

Genomic landscape of divergence/speciation

Genomic regions of statistically unusual high divergence point to regions important for speciation? Riesch et al., Nat. Ecol Evol.

How can we detect speciation loci? Genetic conflict through DMIs/Ecological maladaptation

Challenge 1a: F1 hybrids don’t reproduce – simulation; 1b: some F1 hybrids reproduce – Lindtke & Buerkle 2015, Evolution;

How can we detect speciation loci? Challenge 2 – Ecological malaptation

Timema cristinae stick insect – sympatric melantic – no intermediate phenotypes – PCA on genotypes for population population FHA (n=563)

Long divergence time of color polymorphic regions - BEAST

Long-term maintenance of polymorphism versus speciation

How des speciation work? – natural selection/sexual selection

How does assertive mating affect speciation? Migration/mating/selection

:Lindtke & Yearman 2017 JEB

What do divergence peaks tell us about speciation?

Simon

How much gene flow during speciation?

IM model – Nielsen & Slatkin 2000; Wilkinson-Herbots 2008; IIM model – Wilkinson-Herbots 2012

How does gene flow vary along the genome?

Open questions from a population-genomic point of view

Lessons – demographic first, then selection; Fst outlier scan; islands of what? Sweet spot; background selection, mutation-rate variation

What we need? Joint inference

Amount of effective gene flow varies as a function of slection and recomb.

: Aeschacher & Burger 2014, Genetics

Coalescent theory provides the link to observable variation – structural coalescent

The within-source genetic diversity can be estimated

Migration events reduce ?

Population differentiation decreases with recomb.?

Random allocation of selected loci along the genome

Averaging over the unknown positions of the selected loci

Theory suggests a compound parameter selection density

Example – Mimulus guttatus and its selfing sister species M. nasutus

Gene flow from M. nasutus into M. guttatus – Bardan et al., 2014, PLoS genetics

Selection against introgression maintains a species barrier in sympatry in the South – Aeschhacher et al., 2017, PNAS

Bimodal distribution of pairwise differences for the sympatric Northern species pair – contiguous intronic blocks of size 250 bp

The PMF of the number of pairwise differences can be computed exactly.

IM model provides a bad fit for the sympatric Northern pair

ISC model accommodate the observed distribution … in Northern pair

A long time of complete isolation followed by recent secondary contact

The inferred rate of (effective) gene flow increases with recomb. rates

Fitting a deterministic selection-migration model

Recom. Rates is strongly reduced in gene-poor centrometric regions

Positive correlation betweeon recomb and gene density reverse the signal of BGS

How does gene flow vary along the genome?

A parameter landscape of effective gene flow along the genome

Konrad

Towards model based scans for barriers to gene flow

Litter direct evidence - Cruickschank & Hahn 2015; Noor and Bennet 2009

: Fst outliers in Heliconius are minaly due to reduced pai

The metaphor of a uniform “sea level” of background divergence is silly

:Fst trajectories are hightly stochastic, msprime Kelleher et al., 2016

How to make sense of outliers of divergence?

Heliconius data – H. Melpomene rosina & H. cydno; WGS data, 10 individuals per species from sympatric pops in Panama; autosomal intergenic sequences cut in short (64 bp) blocks – Lohse et al., 2016

Simulations – assumptions – emr 1.9x 10^9 (Keightley et al., 2015); 4 generations per year

How does this help to interpret Fst scans? 19.2 kb windows (300 blocks)

We can do better than Fst!

: Greater power and lower fase + ve rate than Fst

: Fewer outliers (20%)

What about heterogeneity in Ne?

: allowing varying Ne

Why not estimate me locally too?

Allowing for heterogeneity in N3 AND me=0 greatly improves model fit.

Why not estimate me locally?

Quantifying the aggregate effect of selection against migrant alleles