Improving communication for interdisciplinary teams working on storage of digital information in DNA

Interdisciplinary teams make significant advances in life sciences

Intersection between these fields is clearly beneficial. Information theory has already underpinned many advances in life sciences, from adapting Levenshtein coding to create error-correcting molecular barcodes used in multiplexed DNA sequencing⁵ to Burrows-Wheeler transformation of reference genomes implemented in several short read aligners^6–8. A molecular biologist may see the process of storing information in DNA as a very physical process, progressing from DNA synthesis (writing) to amplification (copying) to sequencing (reading). To an information theorist, this is a noisy channel: a series of transformations through which information is transmitted and the outputs observed. Differences in the way experts in these different fields describe their data and results can hinder collaboration and restrict impact. As a result, publications have the potential to be an ineffective hybrid of accepted nomenclature and data presentation within the intersecting fields with few readers, both in the team and outside, able to fully understand the publication as a whole.

Unambiguous communication can be challenging and misunderstandings can pass unnoticed

Unsurprisingly, common nomenclature between the intersecting disciplines has disparate meanings. Use of the word ‘qubit’ can lead you to believe that some DNA needs quantifying⁹ or you may be discussing quantum information or quantum field theory¹⁰. This complicates communication; misunderstandings have the potential to pass unnoticed, only becoming apparent downstream. Examples of such misunderstandings are the use of the words errors, erasures, and substitutions when retrieving data through DNA sequencing. To an information theorist, an ‘error’ refers to a falsely read symbol, for example when an A in the DNA sequence is falsely read as a C, distinct from an insertion or deletion. An ‘erasure’ would be a read that was possibly so uncertain that it is neither called as an A, C, G or T, but distinct from a ‘deletion’ in that the read is not simply missed but we are made aware that there is a missing symbol at this position in the DNA string. An ‘insertion’ is a symbol read, when no symbol should exist. To a molecular biologist and DNA sequencing expert, all of these would be described as read ‘errors’. To them, errors in the information theoretic sense would be called substitutions.

Glossary

We have created a glossary defining basic terms in molecular biology, information theory and computer science etc. that are relevant to DNA-storage, for those unfamiliar with one or more of these disciplines. This proved to be a useful aid in early discussions within our team and helped to identify areas of nomenclature ambiguity which if not addressed may have complicated communication downstream. We have already experienced the advantages of sharing this within our team and with collaborators to facilitate exchange of ideas with them.

Our glossary is held on a cloud storage system, and can be found at https://goo.gl/x6B73Q or https://rebrand.ly/dna-storage-glossary. To allow an open and inclusive discussion of how we might improve communication within this emerging community, we encourage others to critique and contribute to the glossary. The document permits “Suggestions” (proposed edits) and “Comments” to be added, and we will review these regularly and update the document as a resource for our research community.

Controlled vocabulary

Leading on from this, we are developing an evolving controlled vocabulary allowing team members to communicate precisely. This has been particularly beneficial during technical discussions — for instance, to us data packet refers to part of a DNA sequence that decodes to digital information, and excludes parts that are designed to facilitate DNA sequencing or indexing.

Use of a controlled vocabulary is something that the community may wish to agree upon. For example, one question we pose is — what should we name these DNA sequences that encode digital information? Following the practice of genome scientists, we initially called collections of such DNA sequences libraries. However, working with such samples caused confusion with our colleagues in a molecular biology laboratory: in a Next Generation Sequencing context, the term library is commonly used to describe DNA fragments that have been prepared for DNA sequencing. We now propose to refer to DNA sequences that store digital information as inDNA (for ‘information-carrying DNA’). To refer to inDNA prepared for DNA sequencing, we can now unambiguously talk about a library of inDNA.

We would like to invite others to contribute to the development of a controlled vocabulary so that we might be able to communicate more precisely. We have included a few entries within our glossary document.

Presentation can be improved by including a short plain-language summary

The concept of standardising presentation of data and methods is not a novel idea in the life sciences, with ‘minimum information’ standards ensuring that publications contain the information necessary to interpret the experimental data. These are typically technique- or study-specific, e.g. MIAME (microarray experiments)¹², MIQE (quantitative polymerase chain reaction)¹³ and MIFlowCyt (flow cytometry)¹⁴. Such an approach may not be appropriate to publications relating to DNA-storage applications for some time, as these typically encompass a number of disciplines, each with its own established data description standards and many of which use rapidly changing technologies. It is not appropriate or practical to standardise such a diverse range of technologies and disciplines. Rather we should respect the accepted discipline norms, blending these together to permit DNA-storage standards to evolve.

Even publications that sit predominantly within a single discipline may be of interest to those unfamiliar with that discipline and benefit from the inclusion of a whole-paper plain-language summary. As standard with plain-language summaries this should simply report the basic rational, methodology and main findings. Box 1 is a whole-publication plain-language summary of 2 that we have written as an example.

It may also be useful to provide a plain-language summary of a specific technical aspect of a publication. For example, a molecular scientist may not understand the details of a complex mathematical algorithm (and nor should the description be altered specifically to allow them to), but an appreciation of how the output impacts aspects of the project relevant to them may be sufficient. We illustrate this using a paragraph from 4 (from p.5, Methods — Address Design and Encoding). This was read and discussed by the first two co-authors of the present paper, EEH and JS. Figure 1 highlights terms that either EEH, a molecular biologist (purple shading), or JS, an information theorist (yellow shading), found difficult to understand. Joining forces and explaining all terms to each other, they were able to understand the paragraph in depth.

Figure 1. Sample paragraphs from 4.

Terms that may not be clear to non-specialists in particular fields are highlighted in purple and yellow, corresponding to those causing problems for a molecular biologist and an information theorist, respectively. (Used under the Creative Commons Attribution 4.0 International License: http://creativecommons.org/licenses/by/4.0).

As the interdisciplinary field of DNA-storage evolves towards maturity, there will be an increasing requirement for researchers from different backgrounds to understand publications without having access to colleagues from unfamiliar subject areas. This can be achieved in part by including brief summaries, which may make use of our glossary document, in specialised sections of a publications such that they become accessible for researchers from all disciplines.