The Sourdough Project

A major soundscape creation for 2019 is to sonify data for the Sourdough Project. The Rob Dunn Lab at NCSU, the Ben Wolfe Lab at Tufts, and the Noah Fierer Lab at the University of Colorado are collaborating to further the study of microbiomes in sourdough starters. The Sourdough Project has gathered starters from many parts of the world in order to study the bacteria and yeast interactions that create the fermenting acids and leavening gases necessary for the creation of sourdough bread.

In October 2018, the Sourdough Project Team and two artists met at the As If Center in Bakersville NC. The As If Center (Art and science In the field) is the burgeoning vision of Nancy Lowe, who is keenly interested in exploring this fertile collaborative area.The other artist was Ferne Johannssen, freshly graduated from college, and off to see what life outside of Vermont has to offer. Ferne is a visual artist/printmaker. [Interestingly, Ferne made a print on a scoby (symbiotic colony of bacteria and yeast) which grows and ferments in kombucha tea.] The purpose of our meeting was to seed an artistic and scientific direction for sharing the data from the Sourdough Project. All three labs were represented and we spent most of our time sharing information and structuring the research paper that will come from this study. Here is a description of the study from the Rob Dunn Lab website:

There are millions of kinds of bacteria and fungi on Earth. We have found several thousand species in human belly buttons alone. Yet if you mix flour and water, the community of organisms that colonize the resulting concoction is almost always composed of a small handful of organisms that are able to leaven bread, yielding a sourdough starter. How this happens is one of civilizations great mysteries, a mystery at the heart of the bread making (and, for that matter, traditional beer brewing). Yet, while bakers understand how to make starters, the underlying biology of the species in these starters remains mysterious. Starters can produce similar effects on bread (and similar flavors), despite being composed of different species, a key different ingredient. Conversely, starters composed of the same species sometimes yield different flavors. Then there is the issue of what happens to starters over time. The organisms in starters are hypothesized, by some, to stay the same over time—an old growth forest of miniatures—even if their living conditions change. Few ecosystems are so (apparently) stable. Then again, starters can change through time, sometimes suddenly. Starters are, if anything, predictably mysterious. But not for long. We aim to understand the biology underlying the differences among starters and the changes (or lack of change) in starters through time.

The last sentence of this description is what I honed in on. My current sense of how to render data as sound is that it would be most effective with data changes (or lack of) across a timeline. The other word that caught my eye is biology. What is biology? The science of living matter in all forms and phenomena, with special reference to origins, growth, structure, behavior and reproduction. The bases of biology are macromolecules (proteins, lipids, nucleic acids and carbohydrates), cells, and evolutionary changes creating phylogenic families across species. With sourdough starters, we are at the microbial layer of life. On the microbial level, diversity rules and it may have something to teach us. That is what I hope!

The Sourdough Project team had a conference call a few weeks ago, where we saw some of the data analysis of the samples, and received updates from each of the labs. Patterns are starting to emerge as the data is narrowed and focused into categorical relationships. This is the crossroads where it all comes together in the question: What do I want from this data? This most interesting question was posed our first night at As If Center, as we sat around an outdoor fire: what is your currency? what do you want from this project? I can’t remember ever having been asked that before.

The bakers who sent in samples want to know the microbiotic fingerprint of their particular starter. The scientists want to discover some new information about the ecologies of sourdough starters in general. The artists are interested in translation, transposition, representation of the discoveries found in the fingerprints. For myself, I am looking to identify a timeline and voice the bacteria-yeast exchange that is fermentation and leavening. Here is a diagram of a potential time frame:

Water + flour =

aab(acetic acid bacteria) ~ LAB (Lactic Acid Bacteria) ~ Yeasts

which give rise (the timeline but also a phase within the process)

to VOCs (Volatile Organic Compounds) aroma

To my ear this begins with the very lively interaction of the organisms that changes over time into a lighter, gaseous state. There is an alchemy that takes place and we are trying to hear and understand that.

Still looking at TRIC (Terry Riley’s In C) as a template for orchestrating interesting timbral relationships in this context. Pattern 35 is a possible frame for rise which seems to be the name of this piece. Pattern 35 jumps to a start with an eighth note run. This is the organism interaction phase. Then the mid section is where the rise happens with more space and elevation in tone. Then the aromatic texture is very open and light and unfinished.

What other sound elements might lend to this soundscape? There are likely real live sound samples to be had from this process. Another thought is what if each starter could have its own microbiome sounded out? To do this, I need to see more deeply into the data then I have to this date.

More to come…

*Photo from The Sourdough Project website

Sonification and Life Forms II

Since excitedly sharing the results of a sonic analysis of Lemur Gut Microbiomes, I have been working up a soundscape based on the quick sketch included in the first Sonifications and Life Forms post. (You can hear both below.) In order to get feedback on the work, I sent the first blog post to Mark Ballora, with whom I had taken a data sonification workshop in May. His response helped me realize the need to clarify my sonification process. So here is a description of the project:

The purpose of the sonification is to illustrate the changes in baby lemur microbiomes from birth to weaning. Microbiome data was captured through fecal samples taken at birth, through nursing, introduction to solid foods, regular solid foods, and two times while the babies were weaning. The sonification will illustrate changes in the type and amount of bacterial phyla present at each of the six sampling stages for all three lemur babies. In addition, the mother’s microbiome was sampled at the time she gave birth, so her profile, which was assumed not to change, provides a baseline adult profile with which to compare the babies’ changes.

There were 255 strains of bacteria collected over the course of the study. These fell into 95 classes and 35 phylum. I focused on the phylum, as my plan was to assign a note value to each bacterial data point, so I needed a smaller data set. The data set was narrowed further (and made more interesting) by focusing on a family: a mother Pryxis, and her triplets, Carne, Puck and Titan. This group allows us to not only hear the variety of changes in the babies’ microbiomes, but compare the changes as well.

The original data set included 9 lemur babies and 7 mothers. So the first step was to go through the phylum data sheets and pull out the profiles for Pryxis, Carne, Puck and Titan. A phylum profile would be the type and amount of each phyla present at each data collection point. The profile changes over time at each collection point. The microbiomes of these four lemurs housed 15 phylum (at a density of >.001) out of the 35 found in the entire study group.

The next step was to assign a note value to each phyla. Since there are only 13 notes available in the chromatic scale, some phylum would need to be on the same note, albeit a different octave. Same note, different octave will lend a tonal consonance to the profiles. So what might this consonance represent? There were 5 phylum that had the greatest density and presence in all the samples, so I assigned those to the note G from octave 1 to 5. The remaining 10 phylum were assigned note values based on their presence throughout the profiles, and on their consonance/dissonance with the tonal center G.

In order to capture the density of each phyla, a midi velocity range was aligned with the decimal percentage of the phyla in each profile. Midi velocity settings determine the force with which the note is played. Thus the velocity ranges render a clear sense of presence or loudness to each note played. The decimal percentages ran from .001 to 1.0 and the midi velocity range runs from 1 – 127. Here is a chart of how these ranges overlap:

So for example, Protobacteria present at .25473 would be represented by the note G at octave 3 set at 40 velocity. The largest sample in all the data points captured for this project was around .9 and the smallest was .001 (this was a cutoff point as there were bacterial phylum present down to .0001 ranges.) Here is the chart for Titan showing note assignment and density values through each sample stage:

My sounding board for this data comparison is Ableton Live, a digital audio workstation (DAW). The individual lemurs are represented by a “voice”/midi instrument in Ableton. Tuck, Titan and Carne are bell-like voices that blend together, while Pryxis, the mother, is a warm, pervasive woodwind. She envelops and contains the changes in the babies’ phylum profiles.

All lemurs had a Phylum Profile Chart like the one above. In the DAW, the instrument track for Titan, for Puck and for Carne contains a midi-clip of notes of the phylum colonies present at each stage of dietary change, which were then laid out as a “scene” in Ableton. As example, Titan’s Phylum Profile at birth was

Protobacteria (Note value=G3) set at 101/127 in intensity

Euryarcheatae (Note Value=A2) set at 34/127

Firmicutes (Note Value=G2) set at 11/127

Cyanobacteria (Note Value=A#4) set at 1/127

Other Bacteria (Note Value=B3) set at 1/127

Spirochaetae (Note Value=G5) set at 1/127

Titan’s Birth Phylum Profile is the multi octave chord GAA#B. Three of the phylum were barely present, so those tones are almost inaudible in the chord. However, 2 Gs and the A ring out. The total number of phylum present in each dietary stage varied from 3 to 14, so the multi octave chord becomes more dense and dissonant when the phylum are so varied. Here is a look at the tracks (individual lemur voices) and the “scenes” (which are the phylum profiles from all 3 babies at each stage.)

The first sketch was just the mother’s phylum profile droning under the three babies’ profiles expressed as a stacked megachord. All 3 baby profiles rang out together four times at each stage, starting with birth and ending with the second wean. What could be heard was a homogeneity and consonance between the Mother and babies at birth that gradually became more diverse and dissonant as solid food was introduce. However, by the second wean, the babies’ and mother’s profiles become more consonant again. The researcher said this illustrated the conclusions of her study.

As a soundscape artist, I felt there was more here than just that basic chordal movement. The babies’ phylum profiles were quite different from each other as well, which is lost in the chord presentation. For example, Carne’s birth profile has only 3 phylum, while Titan has twice that amount. One way to hear this level of contrast in the baby profiles is to articulate the chords into riffs. Now we can hear the interplay of the changes in their microbiomes. In addition, we can hear how consonnant/dissonant and dense the phylum become as outside food is introduced into their systems. Titan’s phylum profiles arpeggiate down, Puck’s go up and Carne’s go down then up. A practiced deep listener could key in on a particular profile and follow it through to the end. I played around with rhythmic shifts to create more movement in the stages where the phylum profile were incredibly dense and diverse. The last two arpeggiating riffs you will hear are all of the phylum notes sounding through twice. And listen for the elevated levels of Protobacteria in all 3 profiles at birth – that G3 rings out at that point.

As I put this full family profile together, another more nuanced movement in the data appeared. In the chord rendering, I heard the data get more dissonant and dense from nursing through first wean, and then the phylum thinned out and became more consonant at the last wean. In the riff rendering, I can hear a contraction and more consonance at the Intro to Solid Foods stage as well as the second Wean. That was not clear in the chord presentation. When I checked my data records, there was a drop in the number of phylum present between Nurse and Intro stages. I love that a nuance appeared in the listening that made me go back and check the data. That is exactly how I hope this process will work.

Some other things for future consideration:

Aligning each phylum tone to a particular beat might help the listener hear the differences from stage to stage more clearly.

When assigning notes to data points, closer attention to the harmonic overtone series might help clarify the role consonance and dissonance play in hearing the data.

The voices of the baby profiles have similar timbre as a unifying element. The profiles could have very distinct voices which might make the variances in their profiles more audible.

Up next – Sourdough Songs.