[WIP] tiled synthesis simulation

[ericmjl]

Good start! Might need you to walk me through your code in person. Three things that I'm currently checking for - the tiles are overlapping, and that they're in DNA rather than amino acids, and that both the forward and reverse strands are available.

I also found some papers that might give some inspiration:

• Multiplex pairwise assembly
• Dial-out PCR

Give them a read-through, and let's talk about them in the lab (whenever you're in).

[Starnite]

Do you remember what search criteria we used when going through the database? I’m searching for 2011-2016, human H3N2 type A PB2, but I’m getting a lot more results than I downloaded before break.

[ericmjl]

I don't remember the search parameters, but you could try removing duplicate sequences (it's one of the additional options available). Good reminder this time round to record the search params down too. :smile:

[ericmjl]

Great work!

Here's another consideration - I think there's many more unique nucleotide sequences than unique protein sequences. Makes sense, because of the degenerate 3rd codon. If you were to use a reverse translation (AA -> DNA) of the protein sequences vs. straight-up using known nucleotide sequences, what's the difference in the number of unique tiles that have to be synthesized?

[Starnite]

I originally looked for unique nucleotide sequences in the database (there are nearly 3000), but my thinking was that for the most part, the polymorphisms wouldn’t really matter unless it resulted in a translation to a different amino acid. If I did reverse translation, I could either do every possible codon combination, or the most common codon for influenza.

[ericmjl]

True, nucleotide degeneracy in the 3rd codon position might not matter for protein expression, but from an on-chip DNA synthesis standpoint, it very likely will matter. Each degenerate sequence is a new strand that needs to be made, and there's an upper-limit (not sure how much right now, but assumed to be on the order of 10K-100K) of the number of unique nucleotide strands that can be synthesized on one chip.

[Starnite]

I might be misunderstanding something here – wouldn’t there be far too many possibilities if we took into account every degeneracy for every amino acid in the protein?

I will be back tomorrow afternoon – snow got me stuck in NY for an extra day. See you soon!

[ericmjl]

You're right!! I was hoping you'd highlight this independent of myself telling you that point :D.

What would be the engineering implications of this point? If the goal is "make protein sequence (in fragments) while minimizing number of nucleotide fragments required", how would you go about designing the oligos to be synthesized on-chip?

On Sat, Jan 7, 2017 at 11:07 AM, Starnite notifications@github.com wrote:

I might be misunderstanding something here – wouldn’t there be far too many possibilities if we took into account every degeneracy for every amino acid in the protein?

I will be back tomorrow afternoon – snow got me stuck in NY for an extra day. See you soon!

From: Eric Ma notifications@github.com Reply-To: ericmjl/protein-systematic-characterization < reply@reply.github.com> Date: Wednesday, January 4, 2017 at 2:01 PM To: ericmjl/protein-systematic-characterization <protein-systematic- characterization@noreply.github.com> Cc: Jiwei Zhong vivzhong@mit.edu, Author author@noreply.github.com Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

True, nucleotide degeneracy in the 3rd codon position might not matter for protein expression, but from an on-chip DNA synthesis standpoint, it very likely will. Each degenerate sequence is a new strand that needs to be made, and there's an upper-limit (not sure how much right now, but assumed to be on the order of 10K-100K) of the number of unique nucleotide strands that can be synthesized on one chip.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/ ericmjl/protein-systematic-characterization/pull/61#issuecomment-270456187, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ ANrO2aEdcgMrtPeG06KxdQ2MnZmlPshvks5rO-yWgaJpZM4LBsDb.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-271092210, or mute the thread https://github.com/notifications/unsubscribe-auth/ACgnjqqYSxwRnbrUlYjornuRGud05VBxks5rP7hVgaJpZM4LBsDb .

[Starnite]

The two articles I mentioned today:

http://www.nature.com/nature/journal/v432/n7020/full/nature03151.html http://www.nature.com/nbt/journal/v28/n12/full/nbt.1716.html

See you tomorrow!

[ericmjl]

Progress looking good. I think a good next step is to refactor the code into a Python module that you can import elsewhere, and write a few small-scale test cases that can be used to check for correctness in implementation.

[ericmjl]

@Starnite: can you demo your code for me this Thursday, say in the morning? It's so that I can get a better feel for it.

[Starnite]

Sure! Would 10am work for you?

[ericmjl]

Let's do that. What's your calendar email address? MIT or Gmail?

On Tue, Feb 7, 2017 at 8:55 AM, Starnite notifications@github.com wrote:

Sure! Would 10am work for you?

From: Eric Ma notifications@github.com Reply-To: ericmjl/protein-systematic-characterization < reply@reply.github.com> Date: Tuesday, February 7, 2017 at 7:50 AM To: ericmjl/protein-systematic-characterization <protein-systematic- characterization@noreply.github.com> Cc: Jiwei Zhong vivzhong@mit.edu, Mention mention@noreply.github.com Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

@Starnitehttps://github.com/Starnite: can you demo your code for me this Thursday, say in the morning? It's so that I can get a better feel for it.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ ericmjl/protein-systematic-characterization/pull/61#issuecomment-277989803, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ ANrO2RrN0wwEYXAohBJ1cyZm5Kd1iMihks5raGiTgaJpZM4LBsDb.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-278006166, or mute the thread https://github.com/notifications/unsubscribe-auth/ACgnjhYmzGgRgGQc8P37URkRtTnh-sQzks5raHfQgaJpZM4LBsDb .

[Starnite]

Gmail – vivianzhong06@gmail.com

[Starnite]

@ericmjl For 30-bp overlaps and threshold = 1, and testing with 5 sequences to start with, I have 11 re-constructed sequences that are present in the original pool, out of 1896 total.

[ericmjl]

@Starnite Hmm... a bit confused here, how do we start with 5 sequences and end up with 11 reconstructed out of the original 5?

[Starnite]

Many tiles are identical, so there are duplicates of certain sequences – so they’re different paths, but because the nodes are the same, the sequence is the same.

[ericmjl]

Got it. This is interesting, it means that the pool of non-original sequences from the tiles is quite large, under the conditions where one mismatch per tile is allowed. This is good information; it'll elevate the importance of PacBio sequencing + barcoding in rationally arraying the protein variants.

Can you do me a favour and test the code on influenza neuraminidase? Use global H3N2 neuraminidase sequences from the IRD, starting from year 2000 to present. Under the assumption of only perfect matches, how many originals out of the space of total possibles are there?

[Starnite]

Sure, will do!

From: Eric Ma notifications@github.com Reply-To: ericmjl/protein-systematic-characterization reply@reply.github.com Date: Tuesday, March 14, 2017 at 11:28 AM To: ericmjl/protein-systematic-characterization protein-systematic-characterization@noreply.github.com Cc: Vivian Zhong vivzhong@mit.edu, Mention mention@noreply.github.com Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

Got it. This is interesting, it means that the pool of non-original sequences from the tiles is quite large, under the conditions where one mismatch per tile is allowed. This is good information; it'll elevate the importance of PacBio sequencing + barcoding in rationally arraying the protein variants.

Can you do me a favour and test the code on influenza neuraminidase? Use global H3N2 neuraminidase sequences from the IRD, starting from year 2000 to present. Under the assumption of only perfect matches, how many originals out of the space of total possibles are there?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-286457535, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ANrO2b95MzTEERSrquS5bdwuDEALU-Y3ks5rlrIFgaJpZM4LBsDb.

[Starnite]

Hi Eric,

With tile lengths of 210 (30bp overlap on each end included), and testing with 10 sequences, I get 108 original out of 61666 total.

From: Eric Ma notifications@github.com Reply-To: ericmjl/protein-systematic-characterization reply@reply.github.com Date: Tuesday, March 14, 2017 at 11:28 AM To: ericmjl/protein-systematic-characterization protein-systematic-characterization@noreply.github.com Cc: Vivian Zhong vivzhong@mit.edu, Mention mention@noreply.github.com Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

Got it. This is interesting, it means that the pool of non-original sequences from the tiles is quite large, under the conditions where one mismatch per tile is allowed. This is good information; it'll elevate the importance of PacBio sequencing + barcoding in rationally arraying the protein variants.

Can you do me a favour and test the code on influenza neuraminidase? Use global H3N2 neuraminidase sequences from the IRD, starting from year 2000 to present. Under the assumption of only perfect matches, how many originals out of the space of total possibles are there?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-286457535, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ANrO2b95MzTEERSrquS5bdwuDEALU-Y3ks5rlrIFgaJpZM4LBsDb.

[ericmjl]

Great stuff! Now give it a shot on all sequences. If you want to offload the computation to the GPU workstation remotely, go for it!

Btw, if you're using a for-loop in your code, check out the Python package tqdm. It'll give you an estimate for the total loop time needed early on, which you can use to determine whether to offload or not.

Another btw: I think there's a matrix interpretation of path-finding between two nodes, even for directed acyclic graphs (which is what you're constructing). That might be really useful for speeding up computation. Check out this StackExchange question: http://math.stackexchange.com/questions/1890620/finding-path-lengths-by-the-power-of-adjacency-matrix-of-an-undirected-graph.

[Starnite]

Thanks for the tips Eric! Is there a way to run the program in the GPU server without having to re-install all the packages?

[ericmjl]

I think you can export your conda environment as a YAML file, and then use the conda create command to re-create your compute environment on the GPU workstation using the YAML file.

Check out the conda package manager docs: https://conda.io/docs/using/envs.html

On Wed, Mar 15, 2017 at 11:52 AM, Starnite notifications@github.com wrote:

Thanks for the tips Eric! Is there a way to run the program in the GPU server without having to re-install all the packages?

From: Eric Ma notifications@github.com Reply-To: ericmjl/protein-systematic-characterization < reply@reply.github.com> Date: Wednesday, March 15, 2017 at 11:09 AM To: ericmjl/protein-systematic-characterization <protein-systematic- characterization@noreply.github.com> Cc: Vivian Zhong vivzhong@mit.edu, Mention mention@noreply.github.com Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

Great stuff! Now give it a shot on all sequences. If you want to offload the computation to the GPU workstation remotely, go for it!Btw, if you're using a for-loop in your code, check out the Python package tqdm. It'll give you an estimate for the total loop time needed early on, which you can use to determine whether to offload or not.Another btw: I think there's a matrix interpretation of path-finding between two nodes, even for directed acyclic graphs (which is what you're constructing). That might be really useful for speeding up computation. Check out this StackExchange question: http://math.stackexchange.com/questions/1890620/finding- path-lengths-by-the-power-of-adjacency-matrix-of-an-undirected-graph.Sent from Nylas Mail, the best free email app for work

On Mar 14 2017, at 10:42 pm, Starnite notifications@github.com wrote:

Hi Eric,

With tile lengths of 210 (30bp overlap on each end included), and testing with 10 sequences, I get 108 original out of 61666 total.

From: Eric Ma notifications@github.com

Reply-To: ericmjl/protein-systematic-characterization < reply@reply.github.com>

Date: Tuesday, March 14, 2017 at 11:28 AM

To: ericmjl/protein-systematic-characterization <protein-systematic- characterization@noreply.github.com>

Cc: Vivian Zhong vivzhong@mit.edu, Mention mention@noreply.github.com

Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

Got it. This is interesting, it means that the pool of non-original

sequences from the tiles is quite large, under the conditions where one

mismatch per tile is allowed. This is good information; it'll elevate the

importance of PacBio sequencing + barcoding in rationally arraying the

protein variants.

Can you do me a favour and test the code on influenza neuraminidase? Use

global H3N2 neuraminidase sequences from the IRD, starting from year 2000

to present. Under the assumption of only perfect matches, how many

originals out of the space of total possibles are there?

—

You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHubhttps://github.com/ ericmjl/protein-systematic-characterization/pull/61#issuecomment-286457535, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ ANrO2b95MzTEERSrquS5bdwuDEALU-Y3ks5rlrIFgaJpZM4LBsDb.

—You are receiving this because you were mentioned.Reply to this email directly, view it on GitHub, or mute the thread.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ ericmjl/protein-systematic-characterization/pull/61#issuecomment-286771926, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ ANrO2S6KVQ7BXUMOLfvnUVgzh_X0zklYks5rl_8ygaJpZM4LBsDb.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-286786196, or mute the thread https://github.com/notifications/unsubscribe-auth/ACgnjrqUxuwBCepgD6vPuJDjXHW94OjNks5rmAkkgaJpZM4LBsDb .

[Starnite]

Hmm…I’m currently trying to use the ‘scp’ command to copy the yml file to the server, but I get “permission denied”.

[ericmjl]

Hmm.... where are you copying to?

On Thu, Mar 16, 2017 at 10:33 PM, Starnite notifications@github.com wrote:

Hmm…I’m currently trying to use the ‘scp’ command to copy the yml file to the server, but I get “permission denied”.

From: Eric Ma notifications@github.com Reply-To: ericmjl/protein-systematic-characterization < reply@reply.github.com> Date: Wednesday, March 15, 2017 at 11:58 AM To: ericmjl/protein-systematic-characterization <protein-systematic- characterization@noreply.github.com> Cc: Vivian Zhong vivzhong@mit.edu, Mention mention@noreply.github.com Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

I think you can export your conda environment as a YAML file, and then use the conda create command to re-create your compute environment on the GPU workstation using the YAML file.

Check out the conda package manager docs: https://conda.io/docs/using/envs.html

On Wed, Mar 15, 2017 at 11:52 AM, Starnite notifications@github.com wrote:

Thanks for the tips Eric! Is there a way to run the program in the GPU server without having to re-install all the packages?

From: Eric Ma notifications@github.com Reply-To: ericmjl/protein-systematic-characterization < reply@reply.github.com> Date: Wednesday, March 15, 2017 at 11:09 AM To: ericmjl/protein-systematic-characterization <protein-systematic- characterization@noreply.github.com> Cc: Vivian Zhong vivzhong@mit.edu, Mention mention@noreply.github.com

Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

Great stuff! Now give it a shot on all sequences. If you want to offload the computation to the GPU workstation remotely, go for it!Btw, if you're using a for-loop in your code, check out the Python package tqdm. It'll give you an estimate for the total loop time needed early on, which you can use to determine whether to offload or not.Another btw: I think there's a matrix interpretation of path-finding between two nodes, even for directed acyclic graphs (which is what you're constructing). That might be really useful for speeding up computation. Check out this StackExchange question: http://math.stackexchange.com/questions/1890620/finding- path-lengths-by-the-power-of-adjacency-matrix-of-an-undirected-graph.Sent

from Nylas Mail, the best free email app for work

On Mar 14 2017, at 10:42 pm, Starnite notifications@github.com wrote:

Hi Eric,

With tile lengths of 210 (30bp overlap on each end included), and testing with 10 sequences, I get 108 original out of 61666 total.

From: Eric Ma notifications@github.com

Reply-To: ericmjl/protein-systematic-characterization < reply@reply.github.com>

Date: Tuesday, March 14, 2017 at 11:28 AM

To: ericmjl/protein-systematic-characterization <protein-systematic- characterization@noreply.github.com>

Cc: Vivian Zhong vivzhong@mit.edu, Mention mention@noreply.github.com

Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

Got it. This is interesting, it means that the pool of non-original

sequences from the tiles is quite large, under the conditions where one

mismatch per tile is allowed. This is good information; it'll elevate the

importance of PacBio sequencing + barcoding in rationally arraying the

protein variants.

Can you do me a favour and test the code on influenza neuraminidase? Use

global H3N2 neuraminidase sequences from the IRD, starting from year 2000

to present. Under the assumption of only perfect matches, how many

originals out of the space of total possibles are there?

—

You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHubhttps://github.com/ ericmjl/protein-systematic-characterization/pull/61#issuecomment-286457535,

or mute the threadhttps://github.com/notifications/unsubscribe-auth/ ANrO2b95MzTEERSrquS5bdwuDEALU-Y3ks5rlrIFgaJpZM4LBsDb.

—You are receiving this because you were mentioned.Reply to this email directly, view it on GitHub, or mute the thread.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ ericmjl/protein-systematic-characterization/pull/61#issuecomment-286771926,

or mute the threadhttps://github.com/notifications/unsubscribe-auth/ ANrO2S6KVQ7BXUMOLfvnUVgzh_X0zklYks5rl_8ygaJpZM4LBsDb.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ericmjl/protein-systematic-characterization/pull/61# issuecomment-286786196, or mute the thread https://github.com/notifications/unsubscribe-auth/ ACgnjrqUxuwBCepgD6vPuJDjXHW94OjNks5rmAkkgaJpZM4LBsDb .

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ ericmjl/protein-systematic-characterization/pull/61#issuecomment-286788517, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ ANrO2Wzr8ygw1yKhZFljpyo63g27rccbks5rmAq1gaJpZM4LBsDb.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-287250398, or mute the thread https://github.com/notifications/unsubscribe-auth/ACgnjiI63SQWF8qSb3-FSktBZ2I9Q6wCks5rmfD5gaJpZM4LBsDb .

[Starnite]

vivzhong@ubuntu-gpu.myddns.me

[ericmjl]

Let me try resetting your pw.

[ericmjl]

Try again. Your pw is vivianzhong@mit.edu.

To change your pw, after you login, use "passwd vivzhong", and quickly change it.

Let's see if that fixes the permissions error.

Finally, just checking, you aren't scp-ing to the root directory, but your home directory right? The destination should start with "~/".

On Thu, Mar 16, 2017 at 11:00 PM, Eric Ma ericmajinglong@gmail.com wrote:

Let me try resetting your pw.

[Starnite]

Thanks Eric! So I managed to scp the environment.yml file into the home directory, but it seems like I’d still need to do some installing with sudohttp://datascienceguide.github.io/how-to-install-the-python-data-science-stack-on-a-remote-server and pip in order to use the conda command, and I don’t have sudo privileges.

[ericmjl]

Great progress! I think you can use the Anaconda distribution of Python to get around needing to use sudo privileges. On Fri, Mar 17, 2017 at 5:42 PM Starnite notifications@github.com wrote:

Thanks Eric! So I managed to scp the environment.yml file into the home directory, but it seems like I’d still need to do some installing with sudo< http://datascienceguide.github.io/how-to-install-the-python-data-science-stack-on-a-remote-server> and pip in order to use the conda command, and I don’t have sudo privileges.

From: Eric Ma notifications@github.com Reply-To: ericmjl/protein-systematic-characterization < reply@reply.github.com> Date: Thursday, March 16, 2017 at 11:03 PM To: ericmjl/protein-systematic-characterization < protein-systematic-characterization@noreply.github.com> Cc: Vivian Zhong vivzhong@mit.edu, Mention mention@noreply.github.com Subject: Re: [ericmjl/protein-systematic-characterization] [WIP] tiled synthesis simulation (#61)

Try again. Your pw is vivianzhong@mit.edu.

To change your pw, after you login, use "passwd vivzhong", and quickly change it.

Let's see if that fixes the permissions error.

Finally, just checking, you aren't scp-ing to the root directory, but your home directory right? The destination should start with "~/".

On Thu, Mar 16, 2017 at 11:00 PM, Eric Ma ericmajinglong@gmail.com wrote:

Let me try resetting your pw.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub< https://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-287254394>, or mute the thread< https://github.com/notifications/unsubscribe-auth/ANrO2QvVVifemJQH8QczBHZ3bWB0HuJNks5rmff9gaJpZM4LBsDb>.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-287478749, or mute the thread https://github.com/notifications/unsubscribe-auth/ACgnjlIsH3t59lnF5ycI5v7q2G-RBC6iks5rmv5hgaJpZM4LBsDb .

[Starnite]

I have everything working! But the code as is definitely takes far too long to run, so I’ll work on optimizing it. With 50 starting sequences, we get 4337 original and 290651837 new sequences.

[ericmjl]

Good stuff! I notice that your code is particularly useful for estimating the proportion of "desired" sequences that can show up. You might want to do a Monte Carlo walk on the graph, as an alternative in case optimization takes a long time.

[Starnite]

Hey Eric,

Sorry for falling off the grid! I tried to do what I tried to do what I thought would be a simple simplification, but it ended up making the runtime even longer.

My initial method of adding edges was to simply evaluate every pairing of tiles in the digraph as determined by itertools.product. I figured that this was pretty unnecessary, and changed the code so that it only compared two tiles if the second came right after the first in the DNA sequence. This actually resulted in more edges being added, and thus exponentially more paths to search and exponentially longer runtime.

For example, with 2 sequences, 8 tiles each (for the neuraminidase):

First method: num nodes: 16 num edges: 19 original: 2 new: 5 Second method: num nodes: 16 num edges: 24 original: 2 new: 62

The case where every overlap is perfect should result in 28 edges (2 edges per node in a sequence 7 nodes in a sequence (excluding the last one) 2 sequences), so I’m inclined to think that the second method is correct. I’m not sure why the first method lost a few edges, though – perhaps itertools.product isn’t getting everything?

[Starnite]

Addendum: I checked the runtime using the second method for a few different numbers of sequences, and extrapolating from a fitted exponential curve, the code as is would take 68 days to sort through ~2000 sequences. So I’ll definitely look into random walk MC.

[ericmjl]

Hmm... It looks like it'll be worth us meeting up to look through some code. I have a suspicion that you'll need to incorporate more test cases in test suite, but let's chat in person. Got a time that works next week?

[Starnite]

Hi Eric,

Would Thursday 2:30 work for you? I unfortunately have an exam on Wednesday.

I know you mentioned earlier this semester that you wanted to record me carrying out the 96-well polymerase assay; do you have an approximate date in mind for that?

[ericmjl]

@Starnite Thursday's good; I'll be at the Sloan School of Mgmt all day though (there's a 1-day conference going on there), will it be okay with you to walk over there? If not, we can always do a Hangout call after hours too.

[ericmjl]

As for videographing the 96-well plate assay, let me work out some other experimental details; there's still one transfection experiment I'll need your help with as well, but I need to make sure the protocol we have is good to go.

[Starnite]

Hi Eric,

My 6.036 project deadline just got pushed to Monday, so I’d be happy to make the trip. :) What time works for you?

[ericmjl]

@Starnite I took a look at the agenda for the day, it looks like everything ends at 4 pm. Do you think that's a good time for you? I originally thought 2:30 pm, but as it turns out there's only a 15 min. break there -_-", but I think your stuff is worth spending a good hour (or more) on.

[Starnite]

Yup, 4 is good!

On Wed, Apr 5, 2017 at 10:49 PM -0400, "Eric Ma" notifications@github.com<mailto:notifications@github.com> wrote:

@Starnitehttps://github.com/Starnite I took a look at the agenda for the day, it looks like everything ends at 4 pm. Do you think that's a good time for you? I originally thought 2:30 pm, but as it turns out there's only a 15 min. break there -_-", but I think your stuff is worth spending a good hour (or more) on.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ericmjl/protein-systematic-characterization/pull/61#issuecomment-292054374, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ANrO2afNmi9DabemTjSlTTF3cezEfxi3ks5rtFLBgaJpZM4LBsDb.

[ericmjl]

ping @Starnite: as promised yesterday, let's merge this and start a new PR to continue the conversation!