I got two papers with a co-author finished and submitted today.
One of them is based on the segmenter algorithms that I wrote up in this blog (with some improved parameterization): Segmenting noisy signals from nanopores, Segmenting filtered signals, and Segmenting noisy signals revisited.
The other one is an HMM paper that uses the automatic segmentation and an HMM that models the DNA motor behavior to do the same analysis that was done by hand in an earlier paper “Error Rates for Nanopore Discrimination Among Cytosine, Methylcytosine, and Hydroxymethylcytosine Along Individual DNA Strands” in PNAS. The automatic HMM method was more accurate than the hand curation and feature extraction followed by sophisticated machine learning methods like support vector machines and random forests—I think that we were limited before by the biases of the hand curation and feature extraction.
Unfortunately, we were not able to do side-by-side comparison on exactly the same data, as the original data for the PNAS paper was lost in a 2-drive failure on a RAID that had not been properly backed up.
The paper writing did not take much of my time this summer, as my co-author did a lot of the writing and kept me from procrastinating too much.
I’ve also been working on my book, tentatively titled Applied Electronics for Bioengineers, but without a co-author to keep me honest, I’ve been doing a lot of procrastinating. Yesterday I had planned to show the waveform of the gate signal on a nFET being used for pulse-width modulation, to show the “flat spot” that results from feedback through the gate-drain capacitance. I fussed with that for quite a while, but never got a recording I liked the looks of—so ended up getting nothing written on the book. I’ll probably try again tomorrow, but with a backup plan of working on a different section if that one gets too frustrating, or of writing some of the text that will go around the figure, once I get a figure I can use.
The book is currently 215 pages, but that includes the front matter and back matter. The core of the book is only about 188 pages, with 74 figures. I probably need a bunch more figures (schematics, graphs, photos, …), but those are what take the longest to produce. There’s a little over a month left until classes start, but I don’t see myself finishing by then.