Cardona Lab at the MRC Laboratory of Molecular Biology and at the University of Cambridge, United Kingdom

Lab members
Institutional pages: lab at the MRC LMB and at the Department of Physiology Development and Neuroscience (PDN), University of Cambridge.
Lab webpage at MRC LMB
Biography at Pembroke college, Cambridge, UK.

Please contact Albert Cardona to inquire about the possibility of joining the lab. Helpful tips:

What are you interested in working on? How does it fit in the context of present and past projects in the lab? Where do you see yourself in 3 years? Answers to these questions will be very helpful.
Read scientific publications from our lab and from Marta Zlatic's lab.
Read Gilles Laurent's insight piece on "Connectomics: a need for comparative studies", and Winfried Denk et al. on "Structural neurobiology: missing link to a mechanistic understanding of neural computation". Check out also Carandini's "From circuits to behavior: a bridge too far?".
Read Bargmann and Marder's "From the connectome to brain function", Prinz et all. "Similar network activity from disparate circuit parameters", and Marder and Goaillard "Variability, compensation and homeostasis in neuron and network function".
Read Moritz Helmstaedter's "The mutual inspirations of machine learning and neuroscience" and A. Jaegle, Vahid Mehrpour and Nicole Rust's "Visual novelty, curiosity, and intrinsic reward in machine learning and the brain", and follow up with reading of specific case studies such as Dapello et al. "Simulating a primary visual cortex at the front of CNNs improves robustness to image perturbations". Also Zador's "A critique of pure learning and what artificial neural networks can learn from animal brains". And read work that considers how artificial neural networks could be made biologically plausible, like O. Marschall, K. Cho and Cristina Savin's "Evaluating biological plausibility of learning algorithms the lazy way", and also by Pehlevan and Chklovskii (see below).
Read Jonas and Kording's "Could a neuroscientist understand a microprocessor?", and its golden predecessor from Lazebnik, "Can a biologist fix a radio?" (I am a biologist and I can't fix a radio, but I took home lots from this essay).
Read John Platt's "Strong inference" (1964)", which will teach you how to think about research. And so will Sydney Brenner's "Loose ends" column.
Read Rachel Wilson's lab papers on Drosophila neural function. Her work unusually combines rigor, clarity, and timeliness. Specifically recommend the papers on the olfactory system from the early period (2005--2012) which dissect every synapse of the insect antennal lobe. If new to neuroscience, start from the reviews, e.g., her Curr Op Neurobiol and Ann Rev Neurosci pieces.
Read Angus Silver's review on "Neuronal arithmetic".
Read Valentino Braitenberg's "Vehicles: experiments in synthetic psychology", a brief and most insightful book on cybernetics.
If your background is in computer science, statistics or graph theoretic analysis, check out works by Carey E. Priebe and Joshua Vogelstein, for example "A general approach to progressive learning" where they propose representation ensembling as a neuro-inspired approach to design neural networks that can learn more than one mapping, and also "Modern machine learning: partition and vote" where they compare neural networks with random forests, and propose modeling the brain as a partition and vote machine.
If you have a background in neuroscience, read Wolff and Burrows (1995) for a shock on the actual mechanisms of inhibition, on why e.g., a depolarization can be inhibitory, and how axo-axonic inputs are fundamentally different than axo-dendritic ones; "Proprioceptive sensory neurons of a locust leg receive rhythmic presynpatic inhibition during walking".
Read Gerstner et al.'s "Neural dynamics" book (2014), and then Dayan and Abbott's "Theoretical Neuroscience" book (2000). Find works by Hopfield (and his networks). And find bio-inspired and bio-plausible artificial neural network works, such as Pehlevan and Chklovskii's "Neuroscience-inspired online unsupervised learning algorithms: artificial neural networks" and earlier work by Pehlevan and collaborators such as "Blind nonnegative source separation using biological neural networks", and more on biologically plausible neural networks for linear subspace learning, "A hebbian/anti-hebbian neural network for linear subspace learning: A derivation from multidimensional scaling of streaming data". And related works that try to model e.g. cortical circuits with biologically plausible learning rules such as Sacramento et al. "Dendritic cortical microcircuits approximate the backpropagation algorithm", and Lilicrap and Santoro's "Backpropagation through time and the brain". And works that build artificial networks with the anatomical constraints of cortical circuits e.g. Rutishauser, Douglas and Slotine's "Collective stability of networks of winner-take-all circuits" and "Solving Constraint-Satisfaction Problems with Distributed Neocortical-Like Neuronal Networks ".
If you've never read neuroscience as an undergraduate, that's fine, but it would be helpful--to establish a common ground--to read the Kandel book "Principles of neural science" (5th edition) or, more succinct, Liqun Luo's "Principles of neurobiology". Would be helpful as well to dive deep into Russell and Norvig's "Artificial intelligence: a modern approach" (3rd edition).
If you are a biologist in need of developing an intuition for mathematics, read all the blog posts by Kalid Azad at Better Explained. I recommend in particular the posts on complex numbers (imaginary numbers), exponential functions and the number E, and matrices (linear algebra). Math is a beautiful subject, don't let any early shoddy teaching block you from fully utilizing the power of mathematics in your research. On this subject, you might as well read, and try to do the exercises of, Sanjoy Mahajan's "Street fighting mathematics", a wonderful free book that will develop your intuition for quickly approximating something and therefore helping you find out whether you are mostly right about something, which are extremely useful skills all around.
And last but not least, if you'd like to learn about one of the animal models we've worked on the most (the Drosophila larva), learn about what its body can do from the perspective of the modelers, such as Jane Loveless, Konstantinos Lagogiannis and Barbara Webb's "Mechanics of exploration in Drosophila melanogaster", and Pehlevan et al. "Integrative neuromechanics of crawling in D. melanogaster larvae". To a neuroscientist, the biomechanical properties of the body can be surprising: the nervous system doesn't need to specify every movement; instead, merely guide and endow bodily motions with purpose.

Research: experimental connectomics

We chose the larval Drosophila CNS as the organism of study for its numerous advantages, in accessibility (reduced dimensions), outstanding genetic tools for manipulating and monitoring specific neurons, and comparatively extremely low research costs. Almost every neuron in the larval CNS is a unique, identified neuron that can be found from animal to animal, affording experimental reproducibility.

The stereotypy and unique identity of most of the 12,000 larval neurons have been and continue to be key to our success in relating circuit structure to function. Genetic driver lines that label a single pair of left-right homologous identified neurons are pivotal in relating circuit structure, as reconstructed from volume electron microscopy, to observations of circuit function and behaviour: the results of experiments done in one animal can be interpreted with little if any ambiguity on the basis of synaptic circuits mapped in another animal.

Furthermore, we are now able to image with volume EM the complete CNS of animals on which we have performed experiments beforehand, and can therefore investigate the effect of the experimental manipulation on the structure of the circuit under study; we call this approach experimental connectomics. See Valdes Aleman et al. 2020.

The Drosophila larval CNS has turned out to be a fabulously well-posed system in which to observe and manipulate the activity and the connectivity of neurons in the context of fully mapped and mappable brain circuits within comparatively very short time frames.

(Image credits: Philipp Schlegel)

Research: comparative connectomics

With the extraordinary advances in volume electron microscopy techniques, image processing, machine-learning-based circuit mapping, and transcriptomics, it is now possible to analyse at nanometer scale the brains of small animals. We are now setting up microscopes and methods to sample whole brain connectomes from every branch of the phylogenetic tree, aiming first at small species such as the pygmy squid Idiosepius, the dart frog Ranitomeya, and the dwarf gecko Sphaerodactylus or the small chamaelon Brookesia. But any organism with a brain complete and fully functional yet smaller than about one cubic millimeter is now game for whole brain circuit mapping.

(Image credits: click each for the source publication.)

Research: neural circuit structure and function

The computational capabilities of a nervous system are a function of its wiring diagram and the properties of its components, the neurons and glia. Inquires into nervous system function rely on sparsely known or assumed connectivity between neurons or brain areas. Existing circuitry maps detail the connections between very small ensembles of neurons, ranging from pairs to hundreds. This, in contrast with the magnitude of the nervous system of even small model organisms like the fruit fly Drosophila or the zebrafish larva, with over 100,000 neurons. Understanding the nervous system requires far larger wiring diagrams than what we have now, ideally of complete organisms for which experimental manipulation and monitoring of known neural circuits in the context of behavior is feasible.

To make headway into the basics of neural circuit function, we propose to focus on small organisms for which (1) the complete wiring diagram can be obtained within a grant funding period; (2) the complete neural activity over time can be imaged with GCaMP in free or fictive behaviors; (3) each neuron can be genetically manipulated independently and reproducibly; (4) the electrical activity of individual neurons can be recorded with electrophysiology; and (5) high-throughput behavioral assays are routine.

Today and in the near future, only in the larva of Drosophila are all five experimental approaches feasible. Techniques developed for the larva may soon scale to the adult Drosophila, opening the possibility of applying the five approaches outlined above to limb-based locomotion and active flight, as well as to complex behaviors like courtship and learning.

Every day: what we do

We reconstruct neural wiring diagrams from electron microscopy (EM).
We apply graph-theoretic analysis to EM-reconstructed circuitry.
we model neural circuit function on the basis of a known synaptic wiring diagram and then test the predictions of the model experimentally.
We investigate with electrophysiology the properties of individual neurons that participate of EM-reconstructed circuits.
We develop the software CATMAID for the collaborative reconstruction of neural circuits.
We develop novel methods for semi-automatic reconstruction of neurons and circuits, based on computer vision and machine learning techniques.

The reconstruction of the complete CNS of the first instar larva

Starting in December 2012, we are open to proposals from any lab around the world for the reconstruction of their specific area or neuron type of expertise in the Drosophila larval central nervous system (CNS). The ongoing collaborative effort aims at the reconstruction of the complete wiring diagram of the larval CNS. The estimated 10,000 neurons and 3 million synapses will demand the dedication of 50 man-years of work at current rates (we are working towards speeding it up).

A complete electron microscopy image volume of the larval CNS was produced by Richard D. Fetter and the Fly EM Project Team at HHMI Janelia Farm, and consists of about 5,000 serial sections of 50-nm thickness, imaged at 3.8x3.8 nanometers per pixel. Albert Cardona assembled the resulting several hundred thousand image tiles into a coherent volume using the software TrakEM2, powered by the powerful image registration algorithms developed and implemented by Stephan Saalfeld (see publication in Nature Methods).

International collaborators are trained in our lab at Janelia during a visit spanning one to several months, and thereafter access the EM volume of the larval CNS and reconstruct its neurons and synapses online via the web browser, using the software CATMAID.

Participating labs: James D. Truman, Marta Zlatic, Akinao Nose, Michael Pankratz, Simon Sprecher, Ping Shen, Aravi Samuel, Matthias Landgraf, Chris Doe, Matthieu Louis, Andreas Thum, Bertram Gerber, Volker Hartenstein, David Stern, Andreas Thum, Enrique Martin-Blanco, Wesley Grueber, Christian Klämbt, Davi Bock.

Collaborators

Zlatic lab: behavior on the basis of known circuitry.
Truman lab: neuroanatomy of single neurons and identification of single-neuron class GAL4 genetic driver lines.
Richard D. Fetter: generation of EM volumes of Drosophila larval neuropils.
Hess lab: imaging of EM volumes of Drosophila larva.
Funke lab on machine learning approaches to segmenting neuronal arbrors and detecting synapses from electron microscopy volumes.
Akinao Nose, Chris Q. Doe, Matthias Landgraf, Stefan Pulver and Ellie Heckscher on premotor circuit structure and function.
Michael Pankratz on neural circuits for sensing the internal metabolic state and for feeding.
Simon Sprecher on visual circuits.
Volker Hartenstein, Matthieu Louis, Aravi Samuel, Andreas Thum, Wesly Grueber, Bertram Gerber, Greg Jefferis, Liria Masuda-Nakagawa, Lynne Oland and Leslie Tolbert, Christian Klämbt, Peter Soba on mapping the circuits of multiple areas of the fly larval central nervous system.

Prior instances of the Lab

2012-2019: Cardona lab at HHMI Janelia Research Campus.
2008-2011: Cardona lab (mirror link) at the Institute of Neuroinformatics.