Features

The relacs software platform offers the following main features:

Closed-loop experiments

Traditionally, appropriate stimuli for probing a neuron are generated first. Then some days are spent doing experiments with the new stimuli. Afterwards the preliminary data are analyzed offline. Based on the findings the experimental protocol and the stimuli are modified and new experiments are conducted.

Closed-loop experiments. From Benda et al. 2007.

relacs is designed as an framework for closed-loop experiments that may considerably speed up this traditional approach and in addition offers novel experimental possibilities. In a closed-loop experiment a stimulus is presented (1), the resulting response (2) is analyzed immediately (3), and properties of the next stimulus (4) (mean intensity, standard deviation, spectral content...) are adjusted as needed (see figure).

The closed-loop approach is beneficial on many levels, even for traditional experimental paradigms:

• Processed results are displayed immediately in addition to the raw voltage traces and thus give the experimenter valuable quantitative information about the recorded cell during a running experiment. For example, quality of the spike detection, sensitivity or input resistance of the cell, heart rate, etc.
• Probing the neuron with stimuli that are outside its dynamic range can be avoided and thus the yield of a recording is maximized.
• Advancing of individual electrodes can be automated, such that the chance to get a dual recording is increased.

Closed-loop experiments, however, also offer new experimental designs:

• Optimal search for a neuron's receptive field. More ...
• Search for stimuli that drive a neuron in an "optimal" way (firing rate, coding quality, mutual information). More ...
• Find set's of stimulus parameter that result in the same response (iso-response method). More ...

See our review "From response to stimulus: adaptive sampling in sensory physiology" for more details.

Hardware-independent research protocols

Central to relacs are the "research protocols" (RePros). They control the experiment by analysing and visualizing data and generating stimuli. The relacs-core completely hides all specific hardware details behind the scenes. Therefore research protocols can be implemented independently of the hardware used at a particular experimental setup (data acquisition board, attenuator, temperature sensors, motorized micromanipulators). This allows to use the very same research protocols for all your different experimental setups in your lab, no matter what particular hardware is used. Furthermore, this offers the unique possibility to share research protocols with other labs.

Data-analysis library

relacs comes with an extensive set of data-analysis functions. The functions are implemented in C++ to allow fast and memory efficient data-analysis as it is required for closed-loop experiments.

• Basic statistics (mean, standard deviation, median, correlation ...)
• Spectral analysis: power spectrum, transfer function, coherence
• Linear and nonlinear fits (Levenberg-Marquardt and Simplex)
• Peak detection
• Histograms, interpolation
• Stimulus generation: pulse, saw tooth, band-pass filtered white noise, Ornstein-Uhlenbeck noise
• Firing rates: mean, binned, convolved with kernels (rectangle, triangle, gaussian, exponential, ...), 1/ISI
• Interspike intervals: histogram, CV, serial correlation
• Spike timing precision: vector strength, reliability, correlation, synchrony
• ...

New functions are added as required.

Metadata

For offline data-analysis, data management, and data sharing the annotation of the raw data with metadata that specify the stimuli, the recorded neuron, the animal, as well as context of the experiment is necessary. Traditionally, such data have been written into lab books. relacs knows already many of the relevant metadata. In particular, these are properties of the generated stimuli and settings of the controlled hardware. Upon completion of a recording, relacs immediately forces the experimenter to provide additional important information through a freely configurable dialog. In addition, for each recording all configuration files, log files, and settings of the research protocols are saved as well. Thus, relacs minimizes manual efforts as much as possible in providing metadata. This is especially important when it comes to data sharing over public databases like G-Node.

We developed odML, a file format for hierarchically organized key-value pairs that is independent of any specific database-schema and thus can be used as a general file format for exchanging metadata. Through customized terminologies the metadata can be structured and standardized for ensuring interoperability while at the same time not restricting the content, thus providing immediate flexibility. This way, metadata can be directly submitted to a local as well as public data-bases, like for example the LabLog or the German Neuroinformatics Node, without any manual interference. See our paper "A bottom-up approach to data annotation in neurophysiology" for more details on odML.

The NIX project develops generic data models that are tightly linked to meta-data. relacs is using the NIX data model for writing the acquired data and meta-data into a single HD5 file.

Dynamic clamp

The dynamic clamp. During each sampling cycle a voltage V is read in, a current I is computed, and this current is injected back into the cell.

The dynamic clamp is a technique that is used in the neurosciences to artificially introduce ionic conductances to a neuron. This is accomplished by a closed-loop system running on a per-sample basis. First, a voltage value is read in. Then, a current is computed depending on that voltage. Finally, this current is injected back into the cell.

Dynamic clamp can be easily added to relacs by providing a specific implementation of the AnanlogInput and AnalogOutput device classes. Currently, an implementation based on RTAI real-time Linux and comedi for accessing data acquisition boards is provided (DynClampAnalogInput and DynClampAnalogOutput).

Integration of experiment and model simulations

Besides acquiring data from a real experiment, relacs can also run in a simulation mode, where the data are acquired from a model simulation. This is a very important feature, that sets relacs apart from many other data acquisition software:

1. For developing advanced and optimized closed-loop research protocols, the simulation mode is essential to properly develop and test them in advance before they are applied to an experiment.
2. The simulation mode allows to test both, the real neuron and a model, with exactly the same research protocols. This is advantageous for model development close to experimental data.

Models can be implemented as a Model plugin for relacs.

Free software

relacs is distributed as free software ("free" as in "free speech") under the GNU General Public License (GPL).

• Everybody can use the program without the hassle of licenses of commercial software.
• Everybody can modify the program, i.e. add whatever new feature you need directly to the program.
• You can easily share the program and your specific research protocols with your collaborators.
• You always have the possibility to look into source code, so that you know what the program is really doing. This is especially important when it comes to data analysis functions where the science comes in.