| ——Yesterday, march 29, 2008, I woke at 7 in the morning and managed (somehow) to get to the GA/SC Neuroscience Consortium before registration was over.
Over the course of the morning, I met some people, sat through some lectures, and came to some conclusions about the state of neuroscience (at least in this part of the country).
First I’ll give a run down of the lectures.
——The schedule for the day was laid out pretty tight. From 8:30-10:15 was the first general session. I was bleary eyed and muddle headed from waking, but I remember some of the talks well enough to rehash them here.
James Fadel, a name I’ve been vaguely familiar with from our own USC SOM, gave a lecture on Orexin circuits in the mammalian brain. It was a dull start to the day. The interesting bits of information weren’t new to me, and the new bits weren’t interesting.
——After Fadel, we got yet another dose of Orexin (not literally of course, although that may have made the day more interesting).
Rachel Smith (of MUSC), gave a lecture centered on Orexin’s involvement with addiction and extinction of addiction behaviors. She gave some data on their tests involving the Orexin antagonist SB-334867. It was all very behavior centered and sort of vague.
After a couple other lectures, one on a mouse model of Parkinsons and one on Netrin-1’s tole in axonal guidance (both I found pretty obscure and academic), we came to a lecture of more interest to me.
——Sarah Sweitzer, also of the USC SOM, gave the lecture “Modulation of Endogenous Opioid Expression Using Herpes Virus”. “Finally”, I thought, “Someone actually doing something cool!” and, indeed, she was. My understanding isn’t great, but basically a herpes simplex-1 virus was constructed (kindly by Dr. Steve Wilson) that contained the µ-opioid receptor (CDNA?) in the sense direction downstream of the HS-1 promoter. The virus was administered via topical application to a hind paw. The lab successfully showed that 1) this successfully caused an over-expression of the µ-opioid receptor, and 2) that this over expression actually decreased the mice’ apparent perception of both chronic and acute pain. Sarah Sweitzer’s lecture was a great change of pace from the previous talks. Between the genetic modification and immunohistochemistry, it was the first one of the day I could relate to with my own lab experience, and seeing someone using viral vectors for neural engineering like this right here in town was exciting.
——After Sarah Sweitzer, we were treated to an equally fascinating (albeit completely and utterly different) presentation by Stevo Bozinovski from SC State University. I was very pleased to see a computational neuroscientist attending. His lecture, “Delayed Reward Motivated Learning: a Computational Model”, was radically different from anything else given that day. I’m a afraid that a lot of the more wet-ware prejudiced researchers had to fight dozing off during his explanations of synthetic problem solving systems (silly bio-centricists). Bozinovski’s lecture went so far over my head that I’m afraid to even give an outline for fear that I’ll get something wrong. He started by explaining problems the AI community has had for a couple decades with constructing learning systems. As an example, he explained previous attempts at making programs that can learn to navigate mazes in order to find an arbitrary spot. These programs, it has been found, must contain three basic components that are beautifully analogous to corresponding functions in biological brains: integration of sensory data, evaluation of that data, and the execution of decisions made based on that data processing. In the context of a maze, the program wanders randomly within the confines imposed by it’s virtual surroundings, gradually learning the structure of the maze, until it comes upon the designated spot. Upon reintroduction to the maze, the programs entropy of movement is drastically decreased as it goes for where it knows the designated spot to be. According to Bozinovski, this set up is essentially a universal problem solving machine and has been applied to many contexts. One of note was a robot designed to balance a stick. The machine had wheels to move left and right, and could sense the angle of the stick in relation to its top surface. By making the task a conceptual maze where 90 degrees is that sweet spot sought after, and increasingly acute angles ever more aversive, the machine quickly learns to move in a pattern that maintains the stick around 90 degrees.
——On a side note, this reminds me a lot of this paper by Gerald M. Edelman in which a similar program was used to control an actual robot within a RW maze.
——After Bozinovski, there was an hour break. I headed downstairs to the hall of posters, where grad students stood in front of poster-boards containing the densest possible distillation of their work.
——At 11:00, we were herded back up to the auditorium. There were a coupel more boring lectures, mostly behavioral studies on addiction, one on HIV and Malaria in neurons and glial cells, and a lecture by Janet Fisher of USC SOM on the diversity of GABA receptor sub-units and their post-transcriptional editing.
I’m glossing over all of that to get to the lecture I waited the whole day to hear – Joe Z. Tsien‘s presentation on the real time encoding of memory. A great review of his research can be found Here. His lecture focused on the now famous earthquake, elevator, air-puff experiment. In that study, Tsien’s lab constructed a new, very dense, electrode array for taking the recordings of over a hundred neurons in real time. this array was used to monitor the hippocampus of live, freely moving mice while they were submitted to various stressful stimuli – an earthquake, a sudden drop, or a puff of air. By displaying the recordings as a 3d model of activity, it became apparent that a distinct “island” within the abstract 3d space lit up for each of the events. Tsien’s videos of this were spectacular. Interesting, after an event, that same island erratically lit up and died down repeatedly for some time. Tsien explained this with an analogy – when you get off a roller coaster, your brain keeps replaying the scary parts, amplifying them, imagining it over and over again.
From these 3d models of real time memory coding, Tsien has proposed a system similar to the concept of engrams (though I don’t recall him using that term). He calls tis the pyramidal system. Similar experiences have islands located near each other within the abstract space, meaning that many of the same neurons are activated to lay down this memory. A group of neurons may code for the concept of “bed”, others for “mine”, and others for place. Tsien showed this by using pattern recognition to find neurons within the mouse hippocampus that fire selectively when the mouse is, in fact, in bed. Another set of neurons fire only when the mouse is not in bed. The videos of wire head mice running around along with a projection of their neural data lighting up in exquisite correspondence to there surroundings will stay in my mind for some time.
|
Nice post. I really liked it.. Don’t forget to update it regularly. I am looking for new updates dying to read more stuff from you.
— College Research Papers · Feb 22, 10:45 PM · #
I recently came across your blog and have been reading along. I thought I would leave my first comment. I dont know what to say except that I have enjoyed reading.
— College Research Papers · Feb 22, 10:45 PM · #