Pathology Grand Rounds: November 17, 2022

November 18, 2022

Next Generation Immunohistochemistry: How to Build Highly Effective Diagnostic Tests in Anatomic Pathology by Andrew Bellizzi, MD

Information

ID: 9138
To Cite: DCA Citation Guide

Download Transcript

00:00Part so today I'm so happy
00:03to introduce our term policy.
00:05He's from university Iowa notice grant.
00:08Grant Grant Speaker invited by GI pathology
00:11and the middle of the Country Life.
00:13So hard copies here.
00:15We in collaboration invite him to come
00:18here for talk so.
00:21Directly from University Northwestern
00:23University of Chicago in 2004. I'm Diana.
00:28University of Virginia from 2004,
00:312008 and then after that he did
00:34one-year GFL in Ohio University.
00:37After that in 2019,
00:392009 right he moved to Harvard Medical
00:42School for instructor for two years
00:44and then he moved to University Iowa
00:46and then from a senior professor as
00:48a professor and I'm full professor.
00:51And that the policy he published
00:54more than 150 papers and then also
00:56book book chapters and most of the
00:59papers published and which chapters
01:00in GI pathology even though chemistry
01:03and the new tumor and that's why his
01:06research focuses on immuno markers,
01:09you know for for all the diseases
01:12and also he has his research
01:15in neuroendocrine tumor.
01:16He has been joining a lot,
01:18a lot of collaborative studies
01:20you know for H.
01:22OK, for Edge grants,
01:23you know his collaborator for the
01:26studies therefore you are consumer.
01:28That's why you know he involves so many
01:30professional societies especially for cap,
01:33US cap and ASAP for cab.
01:36He has been the even though his
01:39country committee from member,
01:41vice chair and chair and also he's a
01:45international even though his commentary
01:47Society Board of directors and also
01:49involve other lot of committees.
01:51US CAP and cap.
01:53I will not mention all of them.
01:55Or he has.
01:58I have so many publications I
02:00mentioned and I gave so many talks.
02:02Also, you know you watch the talks
02:03everywhere you ask cap cap SP.
02:06So I cannot mention how many hits,
02:08more than 100 pages.
02:10So it's better to ask him to talk so.
02:14So Andrew, test your.
02:16Thank you.
02:17Yes,
02:17yeah.
02:32This is this is literally my
02:35favorite thing to do in the world.
02:39Is to is to interact with folks,
02:42with folks like you.
02:44And I'm so privileged to to have this,
02:48to have this position.
02:50You know, the single most important
02:54take home point from any talk is
02:58that the purpose of this talk is to
03:02make a connection with each of you.
03:07I'm here for you and I
03:11became whatever you want.
03:14If you know me from Twitter,
03:16I became IHC guy.
03:18From from you here,
03:20you know I and this talk is
03:23gonna be about how to become,
03:25how to become an act,
03:26how to become an expert.
03:27And the way to become an expert
03:28is to say that you're an expert.
03:30That's one of the, that's one of the,
03:32that's one of the cheat codes.
03:34Uh, I'm a pretty informal guy,
03:37so I invite everybody to relax
03:40and to try to be present.
03:43And actually the hardest thing for
03:45me as I get going is to be present
03:49because it's it's it's anxiety
03:51provoking to be to be in front of,
03:54in front of all you guys.
03:55One thing that I'll say is it
03:58always astounds me that I have
04:00this platform and that we all
04:03and that we all come together.
04:05Because I had a problematic
04:07relationship with my father,
04:09and if it wasn't for that,
04:12I would probably be in fourth period
04:15teaching chemistry right now.
04:17And you know, I need to do it.
04:19I need to do it better because of that.
04:23Ohh boy. We lost. Somebody's.
04:28Somebody's moving the screen around.
04:31And the the other thing I'll
04:33the other thing I'll say.
04:38I've given this talk multiple times. And.
04:41And the the last time I gave it was at
04:44Mayo and I wanted to and I wanted to say,
04:48but I was so dumbstruck by
04:50what I encountered at Mayo,
04:53which I'll briefly relate to you,
04:55that this talk is, you know, if I gave
05:00the talk that I actually came to give,
05:02which I'm not sure that I will,
05:04this talk is my love letter to
05:07immunohistochemistry and I give lots
05:10of kinds of immunohistochemistry.
05:12Box I give.
05:13I give ones that are love letters.
05:16I give ones that are.
05:17I call myself an immuno optimist and
05:20immuno pessimist and an immune realist.
05:22I give an immuno pessimist talk
05:25this past weekend to the CAP CSA.
05:28The title of the talk was
05:30immunohistochemistry is in the
05:32midst of a quality crisis and low.
05:34Her two is a symptom and most of the
05:37talks that I give are immuno realists
05:39and I give a lot of nuts and bolts.
05:42Didactic.
05:42About diagnostic applications
05:45of immunohistochemistry.
05:50So this is this is the title of
05:52the talk that I submitted next
05:55generation immunohistochemistry how
05:57to build highly effective diagnostic
05:59tests in anatomic pathology.
06:02But I think maybe this is this is
06:04the talk that I'd like to to that I'd
06:07like to try to give simultaneously.
06:09On the invitation and advice of Zuchen,
06:12who actually invited me to give
06:14this talk three years ago,
06:16and he said that my target audience
06:19was trainees and young faculty.
06:22And so how about to the next
06:25generation of pathologists,
06:26how to cultivate a highly successful
06:29career in anatomic pathology?
06:31You, you have the tools,
06:33the most important,
06:35the most important insight that
06:36I can give you is to do this.
06:39You have to know.
06:40Yourself.
06:43So if I give the straight talk,
06:45this is the elevator pitch.
06:48I can build an immunohistochemical
06:50test to virtually any target,
06:53and these 150 publications
06:54are predicated in that,
06:57and the value proposition is for all of you.
07:00What ioffer you is at the end of this
07:03lecture is probably too ambitious,
07:06but my goal is to begin to
07:08provide you the tools to be able
07:10to build an immunohistochemical.
07:12Test to any target every time I go
07:14to a meeting and I've introduced
07:18the new immunohistochemical marker.
07:20The most common question that I
07:23get from from you know world expert
07:26end user diagnostic pathologist
07:27is how did you do this?
07:29Is this commercially,
07:31is this commercially available?
07:33Where do I get this and and the
07:36answer is that there are dozens
07:38if not hundreds of commercially
07:41available antibodies.
07:42To virtually every target,
07:46because science is a $160 billion industry.
07:51I can't remember David where I resourced it,
07:55but Google, Google, I know the NIH.
07:58The NIH, for example,
08:02grants $40 billion of of award every year.
08:08And so there's a good jillion
08:10antibody companies and they make.
08:13Immunohistology,
08:13they make primary antibodies
08:15to everything because some,
08:17somewhere in some corner of the world,
08:20somebody is studying, studying.
08:23And so you can make an
08:25immunohistochemical test at any target.
08:30The biography, you know this is.
08:32This is important because this is.
08:34This is to answer the question
08:36how to cultivate a successful
08:38career in anatomic pathology,
08:40and I'll say more specifically how
08:44to cultivate a successful career in
08:47anatomic pathology in the context of.
08:51Constrained resources.
08:51You know, I've worked at a bunch of
08:55different places and there are very few
08:58exceptional places with the best people
09:00and the best cohorts and the best resources,
09:03the vast majority.
09:05The central mass of places are lovely places.
09:10Iowa is a lovely place,
09:12but even within the constraints,
09:14the constraints of a lovely place
09:17if you are smart and passionate.
09:21And have a point of view.
09:23You can have a wildly successful
09:26career in anatomic pathology.
09:28And So what did I decide to lean into?
09:31I decided to lean into immunohistochemistry.
09:34I started out in GI pathology.
09:36I am a GI pathologist,
09:38but I decided to lean into
09:40immunohistochemistry for so many reasons.
09:42And then another thing Zuchen mentioned is
09:45I'm also a neuroendocrine tumor pathologist.
09:48Every lovely place is.
09:51Exceptional at something or a few things.
09:55And the reason that I'm a
09:57neuroendocrine tumor pathologist
09:58is at the University of Iowa,
10:00the most special group,
10:02the most successful multidisciplinary group,
10:05the most highly funded group is
10:07the neuroendocrine tumor group.
10:09And so if you go to a place and
10:11you want to study pancreas cancer
10:12or you want to study colon cancer,
10:15I've studied all these different
10:17things in different phases of my
10:19career and they're the they're.
10:21See or a B minus group in that tumor type,
10:25it's a it's a losing proposition.
10:27So one of the recommendations is
10:30to survey the scene and identify
10:33the the most talented,
10:35most passionate individuals and
10:37allow yourself with that group.
10:41So I think you've already
10:43gotten a sense of this.
10:44I'm anybody is anybody familiar with
10:48the Myers Briggs personality inventory.
10:52I'm a union psychologist on the side and.
10:57I'm an ENFP and ENFP's
11:01are like functional ADHD.
11:04And so this is how a normal
11:06person tells this story.
11:08And this is how I this is
11:10how I tell the story.
11:11It's I tell lots of stories.
11:13It's very iterative.
11:14I say that my my communication
11:17is at times circumlocutory.
11:19But it always comes around to a point.
11:22So the value proposition and I and
11:24I speak to lots of audiences today,
11:26I'm mostly talking to
11:28diagnostic pathology trainees,
11:30but I also talked to base to
11:32basic scientists quite a bit about
11:35challenges in immunohistochemistry.
11:37And I'll say that everybody
11:39benefits from smart,
11:40technically sound immunohistochemistry
11:42and I'm offering to be your
11:46immunohistochemistry concierge.
11:47I I I warmly invite you to reach out to me.
11:50I'm deeply interested in what.
11:52You're interested in,
11:54I'm deeply interested in helping to
11:56advance your academic careers because
11:59I want to interact with bright,
12:00talented, passionate individuals.
12:02And so this talk is really,
12:04it's directed at everybody.
12:06It's mainly directed at trainees.
12:07I love to interact with local IHC
12:10lab directors because I benefited
12:12so much from interaction with
12:14IHC lab directors when I was
12:16a junior member on the CAP
12:19Immunohistochemistry committee.
12:20I I love interacting with clinical fact.
12:22Faculty,
12:23especially junior faculty who
12:25want to do scholarship but have
12:27limited time and limited resources.
12:30And I have a tried and true
12:32method to have a positive study
12:35using what I call next generation
12:38immunohistochemistry as a function of
12:40what my chair at Brigham told me to be,
12:44which was a translational
12:45molecular pathologist.
12:46And I also love to I love to
12:48talk to scientists and I love
12:50to collaborate with scientists.
12:52For the scientists in the in the
12:54room if you were going to read two
12:56papers about if you were going to
12:59read one paper about immunohistochemistry.
13:01So this is and thank you David
13:03for your work on this.
13:04This is a this is a paper
13:07that discusses validation of
13:09immunohistochemical markers in the
13:11research space and it's all about
13:14diet it's all about specificity of
13:16the antibody now for clinicians
13:19like myself that don't have access to.
13:22Gene editing technologies, for example,
13:24to validate their immunostains.
13:26How do I validate my immunostains?
13:29Clinically,
13:29diagnostically with with
13:31large cohorts of of tumors,
13:34gold standard diagnosis, H&amp;E.
13:36But for basic scientists
13:38this is the way to go.
13:41And then this paper is wonderful.
13:42It's from my friend Angelo Dimarzo
13:45that that really applies many of
13:48the principles in the in the Yulin
13:50paper and does it in a case based.
13:57This is an incredibly important concept
13:59and I talked about this in the in
14:02the teaching session this morning.
14:04This concept of fit for purpose
14:06I and I stated it slightly
14:08different in the unknown session.
14:10I said that immunohistochemical assays are
14:14not interchangeable between laboratories
14:17and the quantity of the analyte,
14:19which is something that we rarely measure,
14:21and David's working on it
14:23and my goal is to take that.
14:25Of the clinical laboratory.
14:26But the quantity of the analyte that
14:30we're measuring by immunohistochemistry
14:32indirectly is very variable among
14:34tumor types that you might use the
14:37same immuno for the example that I
14:39gave this morning was got a got a 3,
14:42there's a tiny amount of got a 3IN
14:44para Feo relative to the amount
14:46of peripheral got it 3IN breast
14:49cancer in urothelial cancer and
14:50so for immunohistochemistry to be
14:53successful and the reason that.
14:55Published studies vary so much is
14:58that immunohistochemical assays
14:59should be conceived of,
15:01optimized and validated to
15:03meet a specific purpose.
15:05A single immunohistochemical assay is
15:08generally poorly positioned to meet
15:11multiple purposes and the the current
15:14her two problem is an is an example.
15:18And here's another wonderful example.
15:20ER clinical IHC assays are optimized.
15:25To select patients for endocrine therapy,
15:28that's how they're optimized and
15:30if you wanted to build an ER
15:33assay to distinguish luminal from
15:35triple negative breast cancer,
15:37it would be a fundamentally
15:39different assay and that's why.
15:41Triple negative breast cancers
15:43by gene expression profiling
15:45sometimes maybe 5% of the time.
15:477% of the time they have ER 1 + 1, two 310%.
15:52This low ER problem that was addressed
15:55in the last ASCO CAP guideline.
15:58The original immunohistochemical assays
16:00for ER were optimized to predict survival,
16:04so they were they were prognostic.
16:05But the the current immunohistochemical
16:08assays optimized for this purpose.
16:11Predict response to endocrine therapy
16:14do pretty good but imperfectly for
16:17this application and then this
16:20application is disappears entirely.
16:26Skip this. Everybody loves.
16:27Everybody loves immunohistochemistry
16:29because it's cheap.
16:30It's widely available.
16:31This is an important point that I want
16:33to emphasize to people that want to
16:36do research in immunohistochemistry
16:37is there is a near endless supply of
16:39commercially available primary antibodies.
16:41You just have to be patient.
16:43The first one might not work,
16:44the second one might not work,
16:46the fifth one might not work,
16:47and then you have to decide
16:49when to cut your losses.
16:50These are the these are the advantages.
16:52These are a lot of the challenges.
16:54Yeah, it's not typically quantitative.
16:56And David, it is work is working on this.
16:59This is from my,
17:00as the Chair of the College of American
17:03Pathologists Immunohistochemistry
17:04Committee responsible for putting out
17:06a national and really international
17:09proficiency testing program.
17:11There's lack of,
17:12there's lack of high quality material
17:14for external quality assessment.
17:16And then as a local IHC lab director,
17:19this impacts my ability to have
17:21high quality control material
17:23and then this is the promise.
17:25We want homogeneous,
17:27controlled material of known
17:29analyte concentration.
17:30And there's a couple different
17:33ways to skin this cat.
17:35David working with my friend
17:38Regan Fulton in Sausalito, CA.
17:41You can combine isagenix cell lines
17:44and you get the quantification
17:46from however you do quantification.
17:48Mass spec, for example.
17:50And then there's another paradigm.
17:52There's a guy called Steve Bogan,
17:54and he has a.
17:56Company called Boston Cell Standards
17:58and he coats glass microbeads with
18:01known concentrations of analyte
18:03and that's traceable to to a
18:05reference standard and I mentioned,
18:08I said the immuno pessimist
18:10talks immunohistochemistry is
18:11in the midst of equality crisis.
18:13It's mainly because of lack of
18:16quantifiable reference material
18:17and we are actively working on
18:19this at the at the at the national
18:22level and this is 1.
18:31As next generation Immuno Histochemistry,
18:34we'll talk about selection, optimization
18:37and validation depending on time.
18:44Presented a dry challenge.
18:45One of my friends on your Roden,
18:49who's a wonderful long, she's also
18:51the director of the IHC lab at Mayo.
18:54She said, Andrew, there's this,
18:56she's a she loves thymoma.
18:59There's this wonderful immunohistochemical
19:00marker that was published in a paper
19:03in MJ Surge Path 10 years ago,
19:05and it's not commercially
19:06available and I want to study it.
19:09But we're stuck, and I and I can
19:12solve her problem by finding many,
19:15many commercially available
19:17antibodies that target my babies.
19:20So I do lots and lots of research
19:23and immunohistochemistry,
19:23and my favorite fun thing to do is make
19:27novel diagnostic markers from from scratch.
19:30And so OTC is in my lab,
19:33the most sensitive Pato
19:35cellular differentiation marker.
19:37The other thing I like to do is repurpose.
19:39Antibodies.
19:40So IGF two which had emerged as a diagnostic
19:44marker for adrenal cortical carcinoma.
19:47I found a novel diagnostic application
19:51in paraganglioma pheochromocytoma packs.
19:54One, we talked about cross
19:56reactivity this morning.
19:57I said polyclonal packs 8 cross reacts
20:00with other packs family transcription
20:02factors including pack 6 and that's
20:04why it's an adequate diagnostic marker
20:07in pancreatic neuroendocrine tumors.
20:09And I was actually having dinner
20:12with Anya at a Kappa HC committee
20:14meeting four years ago.
20:16And she said I want to do a course
20:18about pitfalls and thoracic pathology.
20:20Like, you know,
20:21polyclonal Paxata is positive in thymomas,
20:24but monoclonal paxata is negative.
20:26And she said that.
20:27And I,
20:28from the example that I knew
20:30from PAX 5 and pack six,
20:32I knew that that there was some other
20:35PAX family transcription factor
20:37that was expressed in the thymus.
20:40And I all you have to do is have pub,
20:43pub,
20:43Med and a few bioinformatics tools
20:45that I will share with you to
20:48be able to solve this in silico.
20:51And then doing the clinical study
20:53is a formality angler one is a
20:55diagnostic marker for chondromyxoid
20:57fibroma and it's based on the the
21:00defining molecular genetic event
21:01and that lesion hey one and then
21:05specifically hay one immunohistochemistry
21:07directed to the end terminus of the hay one.
21:11Gene or I should say protein as a
21:14diagnostic marker for mesenchymal
21:16chondrosarcoma where a diagnostic
21:18marker did not cry previously exist.
21:22And then the last one got a four,
21:24which I've brought up as a pan
21:28GI marker expressed in the upper
21:30tract and not in the colon.
21:31So very complementary to SAP B2,
21:34which is more of a lower GI colon
21:36specific marker and I don't know
21:38that we'll get to all these
21:39specific examples and that's fine.
21:41The concepts more important than
21:43the than the detail.
21:46So next generation immunohistochemistry,
21:48this is my, this is my stick.
21:52And this term I, I sort of Co developed
21:56with with my very close friend that I
21:59worked with incredibly closely at Brigham.
22:02Jason hornick. Jason and I do next
22:05generation immunohistochemistry.
22:06He does it almost exclusively in
22:08sarcoma and I do it in pan cancer
22:11diagnostic applications because I'm
22:12at Iowa right because he's at Brigham
22:15and he could do next generation of.
22:17Sarcoma and Iowa has a lovely
22:20sarcoma program,
22:21but it does not have an exceptional one.
22:23And so I do next generation
22:27immunohistochemistry and I'm not going
22:28to restrict myself to any tumor type.
22:30I'm going to pick off anything
22:33that anything that I can.
22:35And how do I find markers?
22:37And I had lovely discussions this morning
22:41about my passions for developmental biology.
22:44I spend many Saturdays in the library
22:47reading developmental biology papers,
22:49trying to understand.
22:50First thing I told you guys this morning
22:53is tumors recapitulate cells and tissues
22:55native to the organ in which they arise.
22:58And so it behooves me to understand
23:01the developmental biology of those
23:03cells and tissues and from them.
23:05I derive my diagnostic markers those
23:07are my favorite are are querying the
23:10developmental biology literature.
23:12We also do protein correlates of
23:14molecular genetic events and you
23:16know we did we talked about TP53
23:18and P53 immunohistochemistry RV one
23:21and RV immunohistochemistry this
23:23morning is exemplars.
23:25And then you can also identify
23:27markers identified by gene expression
23:29profiling and I say that's cheating
23:31and the reason I say it's cheating is.
23:34Well, I can't do it.
23:35I don't have the.
23:36I don't have the resources to
23:38do gene expression profiling.
23:39I would rather use my brain to identify a
23:42target than have something come up as the,
23:45you know the the most differentially
23:47expressed marker in some in some experiment
23:50though I have used markers identify,
23:52I have used that strategy
23:55to identify markers.
23:56You know,
23:57this is me being an immuno optimist.
23:58Our diagnostic markers keep getting better
24:01and better and better and you know this.
24:05I'm a sort of a goodwill hunting pathologist.
24:07You wasted
24:10$150,000 on an education or some some
24:14fancy experiments for an education.
24:16You could have got for $1.50 in
24:19late fees at the Public Library.
24:21So this is how I find diagnostic
24:24markers and this is and I'll
24:25say you know this is an example.
24:27This is How I Met cute got a 3.
24:31So this was the 1st paper that was
24:33published in the diagnostic pathology
24:35space that identified got a 3 as a
24:38urothelial cancer diagnostic marker and
24:40this used gene expression profiling.
24:42This came out of Stanford.
24:44John Higgins was the first author.
24:46This is Rob West and Matt
24:49Vander Reines work and they did.
24:51Gene expression profiling of Gu tumor,
24:53so they did kidney,
24:56black bladder and and prostate and
24:57got a three was identified as a
25:00highly differentially expressed gene
25:01in those tumor types and then you
25:04validate it with immunohistochemistry.
25:08Here's another one.
25:10And the defining molecular genetic
25:12event of a tumor is identified
25:15through and through whole genome
25:17or whole exome sequencing, Jason.
25:20And I say the tumor type has been solved
25:23and when the tumor type has been solved,
25:26this is such fertile ground to
25:28identify a diagnostic marker.
25:30So, so, so this defining molecular
25:33genetic event of solitary fibrous
25:35tumor was described and within
25:38a few months Jason and Chris.
25:41Chris. Butcher and Leona Doyle,
25:43wonderful junior now middle faculty,
25:46used to be my resident and fellow
25:48can published a diagnostic study,
25:50and when the immunohistochemical
25:52markers are identified in this fashion,
25:55they are so damn specific.
25:58They're so damn sensitive and specific.
26:00So this is how stat six came to be.
26:04And then I can take it to my lab.
26:06So this was a case,
26:08a spinal lesion that had a little
26:10bit of carrot and positivity that
26:12sat in the file for 10 years
26:14diagnosed as synovial sarcoma.
26:16I had identified this case.
26:17I was doing a TLE one validation
26:20several years ago and it was
26:22TLE one negative and I looked at
26:24this tumor and I actually didn't
26:25think it was synovial sarcoma.
26:27I thought it was a poorly differentiated
26:30solitary fibrous tumor and in fact it is
26:32and I and I say using next generation.
26:35Immunohistochemistry.
26:35You can elevate the ability of
26:40already strong morphologist,
26:42already strong diagnostic
26:43surgical pathologists.
26:44You give them incredibly sharp
26:46tools and they can become
26:48exceptional diagnostically.
26:52And then hit my love letter to
26:56transcription factors history.
26:58Historically, our diagnostic armamentarium
27:00was geared toward cytoplasmic and
27:03membranous differentiation markers
27:05that were empirically discovered.
27:07We would like mince up some tissue
27:09and put it in A and put it in a
27:12rabbit and see and see what's stuck.
27:15And so, you know,
27:16here's a breast cancer with mammaglobin
27:18and gross cystic disease fluid protein.
27:21And remember this,
27:22inside of plasmic differentiation markers,
27:25their expression tends to be
27:27markedly reduced in poorly
27:29differentiated tumors as opposed to
27:31lineage restricted transcription
27:33factors that tend to not always,
27:35but tend to be diffusely strongly
27:38expressed independent of the
27:39differentiation state status of the marker.
27:42And I like to coin a lot of terms
27:44and this is what I call the primacy.
27:47Primacy either means first or best.
27:49So this is for me the primacy because best.
27:51Primacy of lineage restricted
27:54transcription factors.
27:55And immuno optimists,
27:57immuno pessimists and immuno realists,
27:59and really we should all be immuno realist.
28:03Every new immunohistochemical
28:04test is like a new technology.
28:08This you may be familiar with is
28:10the Gartner Hype cycle that's used
28:13to describe new technologies.
28:14When new technologies are described,
28:17we're incredibly enthusiastic about them
28:19and overly enthusiastic about them.
28:21And then they start failing a
28:23bit and then we keep studying.
28:25Studying and studying and we've learned more,
28:27and we learn advantages,
28:29disadvantages, limitations,
28:30pitfalls,
28:30and we end up with this plateau
28:33of productivity.
28:34And I will say the reason that many
28:39immunohistochemical tests historically
28:41or new analytes have suffered from
28:44this is from poor experimental design.
28:48Immunohistochemical when you're introducing
28:50a new diagnostic marker into the space,
28:53it behooves one to examine.
28:56All the lesions in the
28:58differential diagnosis,
28:59if it behooves one to look for potential
29:03additional diagnostic utilities,
29:04and you can do that by staining
29:06a bunch of cases.
29:07But you can also use publicly
29:10available gene expression data
29:12to do the experiment in silico,
29:14and then and then to confirm it with
29:16parents paraffin and this is these
29:18are all papers published by Marku
29:20Marku Miettinen, who's a personal hero,
29:23and he bombs every new.
29:26Immunohistochemical marker,
29:27he studies them in thousands
29:29and thousands of tumors.
29:31So you know,
29:32got a 3 and 2500 tumors 4 for example.
29:37And again, exceptional,
29:39lovely, you know,
29:42doing what you can within the constraints
29:45of the resources at your disposal.
29:47And So what I've done at Iowa over the
29:50last 11 years is do what Marku did.
29:53And so I've developed tissue microarrays.
29:55I, I,
29:56you know,
29:56boasted this morning that I have hundreds
29:59of tissue microarrays of thousands
30:01of tumors of every, of every type.
30:03And I'm now able to do,
30:06able to do this.
30:09And so instead of doing this,
30:12we start at the plateau of productivity.
30:16We start with an informed
30:18new diagnostic marker.
30:24So stains I love most are oligo
30:27specific transcription factors.
30:29That, and I think a lot of people
30:31expect to transcription factor to
30:33have a single diagnostic application.
30:36And I think that's entirely unrealistic
30:38that the body is too efficient to
30:42use a single transcription factor
30:44for a single application it it
30:47reuses things you know and and so.
30:50It just behooves us to know the
30:52different diagnostic applications
30:53for all these different markers.
30:56And so I say stains, I lovers,
30:58Swiss army knives and sappy 2 is, you know,
31:01an example that I'll share with you.
31:03You know,
31:04my approach to every diagnostic marker,
31:07mainly transcription factors,
31:09is I dig very deeply into the basic
31:12science underpinning the marker,
31:15especially the developmental biology because
31:17that's my hop hobby and so sappy too.
31:21You know,
31:21pathologists learned about it about
31:2310 years ago because it was actually
31:25identified as a colon cancer marker in the
31:28context of the human protein Atlas effort.
31:30But every new diagnostic
31:32marker if you go to Pub Med,
31:35has a rich,
31:36generally decades long history in the
31:39developmental biology literature.
31:41And so even though we met SAT B2 in 2011,
31:45this was this was studied by
31:47developmental biologists that's
31:49involved in neuronal development.
31:51Craniofacial development
31:53and skeletal development.
31:54And from the developmental
31:56biology of the marker,
31:58you can infer diagnostic applications
32:00and then you can test them.
32:02You can design the study to test them.
32:05And so here's sat B2 doing the
32:08most familiar SAT B2 thing.
32:10This is a mucinous adenocarcinoma
32:12that before I had sat B2 I would
32:14have said had a lower GI phenotype.
32:16It's homogeneous.
32:17I say homogeneous for diffuse,
32:19strong for CK20 and CDX 2.
32:23And I would have said adenocarcinoma with
32:26mucinous features and lower GI phenotype.
32:28And this one's actually sappy,
32:30too negative,
32:31and this is actually an ampullary
32:33adenocarcinoma.
32:34So this is SATB 2 conferring
32:37lower GI specificity.
32:40Here's another diagnostic
32:41application of SAT V2.
32:43This is an undifferentiated
32:46undifferentiated carcinoma.
32:47Here's a carrot and in the
32:49colon it's MLH 1 deficient.
32:51So this is a this is a not quite a medullary.
32:56It's an undifferentiated MSI high
32:58cancer and it's CDX two negative
33:00and MSI high colon cancers are
33:03often CDX two negative because CDX
33:052 actually has micro satellites in
33:08its exons and is susceptible to.
33:10Frameshift mutation and then
33:12nonsense mediated decay so that it's.
33:16It also explains why the tumors
33:19are undifferentiated.
33:20And here's sappy 2 in this case.
33:22So SAP two has an additional diagnostic
33:25application as an increased it adds
33:29increased sensitivity to the detection
33:32of undifferentiated colon cancer.
33:34Here's another one.
33:35This Jason did because he
33:37was positioned to do it.
33:38He's a sarcoma pathologist.
33:40He did it with my resident Jay
33:43Connor who's now the chief of
33:45GI at at at Sinai in Toronto.
33:48Is this so this was age-old question.
33:51Is this osteoid or is it just
33:54hyalinized stroma and sappy 2 is an
33:58osteoblast differentiation marker.
34:00We knew that from the developmental biology
34:02literature and now this age-old question.
34:04I can say emphatically yes,
34:06this is an Austin.
34:08This is osteoid.
34:09Here's another diagnostic application
34:11and this is where I had the opportunity
34:13to contribute to literature because
34:15this is where I was the best position.
34:18I was best positioned to study this
34:20marker in the neuroendocrine space.
34:22And I had presented this morning that
34:25SAT B2 is a wonderful marker of rectal,
34:28also expressed by appendiceal
34:30well differentiated neuroendocrine
34:31tumors and it's not expressed by
34:35pancreatic neuroendocrine tumors.
34:36And then I had also shared that
34:39this is important.
34:39Because a lot of the transcriptional
34:42machinery that drives pancreas net
34:44development namely Pack 6 and islet
34:47one that transcriptional machinery is
34:49also operant in rectal neuroendocrine tumors.
34:52And so this,
34:53this gave me a tool to clearly distinguish
34:56pancreatic and rectal neuroendocrine tumors.
34:59And then I showed you the morphology,
35:01right.
35:01Everybody can everybody after this
35:04morning's discussion could tell me that
35:07this is type B trabecular paper clip rectal.
35:10And then here's here's another diagnostic
35:13application I shared this this morning.
35:15How about PD neck?
35:16And in PD neck,
35:18sat B2 is a reasonably sensitive,
35:21highly specific marker of cutaneous
35:24neuroendocrine carcinoma,
35:25Merkel cell carcinoma.
35:30And, and there's more, right.
35:32So here I'm being like a, you know,
35:34immunohistochemistry used car salesman.
35:36There's one more diagnostic application.
35:38This is, this is a, this is an
35:41undifferentiated round cell sarcoma.
35:43That's Ewing sarcoma. Ish.
35:46It's a little too pleomorphic to be a
35:49Ewing sarcoma and it's CD 99 negative and
35:53before I had the opportunity to validate
35:56becor immunohistochemistry in my lab.
36:00Because it was hard to come across
36:0210B core rearranged sarcomas,
36:03I use SAP B2 as a surrogate for B
36:07core rearranged sarcomas because about
36:1070% of B core rearranged sarcomas
36:12are strongly sappy, too positive.
36:14This was demonstrated by Christina
36:16Antonescu in gene expression profiling
36:19experiments and sort of tucked in,
36:21you know, as a supplementary fig figure
36:24in one of her papers from 10 years ago.
36:27All right.
36:28So I always ask will it Swiss army
36:30knife and and the answer is yes.
36:32Yes I want more more value with fewer mark,
36:35fewer markers.
36:37All right.
36:37So that's next generation
36:41immunohistochemistry onto optimization.
36:44It's maybe we could say selection
36:47optimization and validation and so,
36:49so there's Twitter, so am I HC guy.
36:53And the reason that I'm IHC guy is
36:56I'm passionate about being connected
36:58with people and having the opportunity
37:01to share things that I'm passionate
37:04about mainly immunohistochemistry.
37:06I'm kind of a nerd and and so I use
37:11this platform to to mentor IHC lab.
37:14Directorship and different directors
37:15are at different levels and I give
37:18different directors different advice.
37:20So this is my friend Frank Ingram,
37:23who's Chuck Town MG,
37:25very popular on Twitter and he took
37:29over as a local icy lab director in
37:33his community practice in Charlotte,
37:36NC and he had no training in in
37:40Immunohistochemistry Lab directorship.
37:41Just like me, I had no formal.
37:44Training in Immunohistochemistry
37:45lab directorship and he said I
37:48want to bring up God up, 3 packs,
37:508 socks,
37:5110 and erg to which I said Hallelujah,
37:54like you must have those markers.
37:56And I provide you know I I talk
37:58a lot right and I write a lot.
38:00So I provided him very detailed
38:03instructions about how I would go about
38:07value optimizing and then validating
38:09what I consider to be pretty routine.
38:12Tried and true robust.
38:16Immunohistochemical markers.
38:18And then here's another one of my
38:19close friends. This is Phil Reese.
38:21Phil Reese is operating at
38:22A at a much higher level.
38:24He's an IC lab.
38:26He's the IHC lab director
38:28at OHSU in Portland.
38:30He took the lab over from Megan Troxell,
38:32who's a Jedi Master of
38:35Immunohistochemistry lab directorship.
38:37And she and I have taught each other
38:39a tremendous amount over the years.
38:41And Phil was looking for something very,
38:44very specific.
38:46He wanted immunohistochemistry.
38:48For Toxoplasma, which is a really,
38:51really niche application and I said,
38:54well,
38:55I certainly don't do
38:57immunohistochemistry for Toxoplasma.
38:59But the one person in the world
39:01that's the best person to talk to you
39:05about Toxoplasma immunohistochemistry
39:06is an unfortunately he passed
39:09away about 18 months ago.
39:11But Sharif Zaki, who was the chief of
39:15IBD pathology at the NIH and I connected.
39:18Fill with Sharif and Phil got what he needed.
39:23So antibody selection you know I I'll
39:26I'll say like all these examples that
39:28I might get to one or two of your
39:31got a three and your end terminus.
39:34Hey one you know I'm selecting
39:36antibodies I'm I'm trying to invent
39:39new antibodies but but Chuck Frank,
39:42Chuck Town, Frank knows from from you
39:44know going to be going to meetings
39:47listening to listening to lectures
39:49what the you know cutting edge what
39:51the advancing edge of diagnostic.
39:53Pathology is and so that's
39:55antibody selection.
39:56What what immunohistochemical
39:57marker do I want to bring up in my
40:02lab for a diagnostic application
40:04or for research application?
40:06And then this is sort of the sequence
40:09of events from maybe from easy to
40:11hard that I go through and actually.
40:14There's a few tricks of the trade that
40:16I learned from David Rim on this list.
40:18So so the easiest thing is to phone
40:20a friend and you can always phone
40:22you can always phone me phone a
40:24friend works great for esoteric
40:26antibodies so connecting fill the
40:28Sharif was was really that was the
40:30one step solution to that.
40:32The published literature is great is great.
40:35I often go to the public I just go to pubbed.
40:39Has there been a been a paper
40:42published with this marker.
40:44What are the figures look like,
40:45you know,
40:45what are the,
40:46what are the pictures look like?
40:47Does this look good?
40:48Does this look like an
40:50immunohistochemical stain of high
40:51quality that I would want in my lab?
40:53And then Nordic QC is the Scandinavian
40:56equivalent of the CAP Immunohistochemistry
40:59committee and they published their very
41:03transparent and they published very
41:06detailed platform specific protocols
41:08for what I refer to as diagnostic
41:11workhorses so that they've got published.
41:14Protocols and usually 10 different published
41:17protocols that are again platform specific.
41:19So you're running a doco auto
41:21Stainer link 48 or you're running
41:23a VENTANA benchmark ultra,
41:25or you have a or you have a doco
41:28omnis or you have a like a Bond 3.
41:30For for about 150 to 200
41:34different diagnostic antibodies.
41:36So that's where I pointed frank for
41:38packs 8 and socks 10 and erg and then and
41:42then these are more shots in the dark.
41:44So these are these are databases that
41:46you can I usually end up when I'm
41:48bringing up a brand new diagnostic
41:50marker I end up with these databases.
41:53Biocompare is pretty dumb.
41:54It just spits out it just spits out a
41:58list a list of vendors and you can click on.
42:01You know,
42:01you can click on the link and go
42:03to the website,
42:04but these are two that I learned
42:06from David actually,
42:08antibody pedia and bench and bench side.
42:11And I I think I like benci the
42:14most because bentsi does for
42:16me it uses AI and does for
42:19me what I used to do,
42:21which is go through every paper
42:23published and say look at the method
42:25and is there an immunohistochemical
42:27method and what's the clone and
42:30aggregate all this information.
42:31And bench side does it for you.
42:34So this is what Nordic QC looks
42:36like and these are examples,
42:38so these are examples of these are all the
42:41different like recommendable her to assays.
42:46And you know it's not just the clone,
42:48it's all the conditions of the test and
42:51you click the link to the PDF and that's
42:54a starting point for your optimization.
42:56And then here's biocompare
42:59antibody pedia ads.
43:02Biocompare just gives you a list of
43:03a list of vendors and you can click
43:05links and go to their websites.
43:07Antibody Pedia will vouch on and
43:09you're looking for you're looking for.
43:12Green Circles will vouch for the quality
43:15of the experiments that were done to
43:19validate the specificity of the antibody.
43:21So all the way back to the beginning
43:23when I said there's these two
43:25papers for the basic scientists,
43:27the one that Angelo de Marzo published,
43:30this case based, and the Eulen 1.
43:32That that David collaborated on and so many.
43:35And then if you do it for a while,
43:37you know some vendors are more scrupulous
43:40with their science than others,
43:42but this gives you a check.
43:43And so we're looking for a green circle that
43:46is referring to an enhanced by validation.
43:49So.
43:50So they're using some of these
43:52techniques like like like you know,
43:55either crisper cast and not to knock it
43:57out or they're using small interfering
43:59RNA to knock it down to validate.
44:02The specificity of the antibody.
44:05And then here is bench side,
44:07which gives me which links to the papers
44:10themselves and not just the papers,
44:12but links to the figures of the papers.
44:15So this is I used to use
44:17biocompare and thank you David.
44:19Now I go to Bankside 1st and
44:21antibody pedia and benzite,
44:23there's they're free for you,
44:25I think you have.
44:26They're free to academicians.
44:27You have to like create a login, but totally,
44:30totally free, incredibly powerful tools.
44:33We're doing on time.
44:35We're doing OK.
44:36Ethical sourcing of primary antibodies.
44:38I mentioned this briefly in
44:40a few conversations.
44:41I will not source an antibody
44:44from Santa Cruz in my lab because
44:48they committed gross animal abuse
44:52and they actually they they they
44:56lost their license to sell.
45:00Animal products derived from animals
45:02and and the loophole is that mice
45:06don't qualify as animals so they so
45:08they can't sell polyclonal's anymore.
45:10They can't sell rabbit,
45:11rabbit polyclonal or goat polyclonal
45:13because those those are considered
45:15animals by the by the government.
45:17But mice aren't considered animals
45:19and so Santa Cruz is still in
45:22business but I won't source an
45:24antibody from Santa Cruz.
45:25Optimization you know I'll I'll sort
45:27of move along because I want to get
45:30to at least one specific example.
45:32Optimization is the series
45:34of experiments that we do to
45:36optimize the signal to noise ratio.
45:39And the way that I do optimization is
45:42is is very guest check revised and
45:45visually oriented and David has a awesome
45:48scientific way to do optimization.
45:50He optimizes the signal to noise
45:52ratio but he he takes the human.
45:55Right out of the equation.
45:57And this here's an example,
45:59you know these are the the
46:02iterative specific experiments that
46:04I perform to optimize a primary,
46:06new primary antibody.
46:07We mainly attend to the
46:10primary antibody dilution,
46:11the type of antigen retrieval and then the
46:15primary antibody duration of incubation,
46:18the choice of adding amplification to
46:22detection or just straight detection
46:24and then the duration of detection.
46:28And then there's validation.
46:29I talked about validation in
46:30the research space.
46:31This is validation in the clinical space.
46:33So an IHC lab director is going to
46:36pay attention to this White Paper,
46:38principles of analytic validation
46:40of immunohistochemical assays that
46:42specifies the quantity and quality
46:45of the cohorts that one should put
46:47together to to to validate an antibody,
46:50which is necessary to put it
46:54into clinical use.
46:56Here's validation in the research setting.
46:58These are the different types of experiments.
47:01And then let's, let's not do mine, let's do,
47:05let's do annyas, let's do the dry,
47:07let's do the dry challenge.
47:08So Anya wanted proteasome subunit
47:12beta 5T and I said I'm she said
47:15this is not commercially available.
47:17I am so interested in this analyte
47:19and it's not commercially available
47:21and I'm devastated.
47:23And I said I'm sure that I'm sure that
47:26I can make an immunohistochemical test
47:29to proteasome subunit beta 5T and.
47:31And we talked about it.
47:32I was giving I,
47:34I Co direct a week long immunohistochemistry
47:37course and these are three of my
47:40lovely faculty and this is this is
47:42what we talked about at the at the
47:44dinner table in Montreal this summer.
47:46And so here's and then you just have
47:49to trace it and I call this like
47:52immunohistochemistry Sudoku right.
47:53And I want to I want to get to the
47:55solution and as few steps as possible.
47:58So all she told me was proteosome
48:00subunit beta 5T.
48:01And I found this paper that she
48:03was referring to.
48:04So it's a Japanese group
48:07and they created their own,
48:09I believe polyclonal antibody to to
48:13this proteosome subunit beta 5T.
48:16Fascinatingly, you know proteasomes,
48:18they're they're cool, they have,
48:21they have lids,
48:22they have a cylinder body and they
48:25have bottoms, they have subunits.
48:26And the thymus has its a unique proteosome,
48:30totally different than the proteosome
48:31and the whole rest of the body.
48:33Because in the thymus,
48:35you need to use the proteosome
48:37to process antigens, to present,
48:39to grow up, to grow up T cells.
48:41So that's the,
48:42that's the science behind this thing.
48:44And then the first thing that
48:46I had to solve is, well,
48:48I needed to know what the gene was and
48:50it wasn't clear what the gene was,
48:52but just Googling around I found
48:55a couple different.
48:56So there's PSMB 5.
48:58And then I started and then it turned
49:02out to actually be PSMB 11 was was
49:06the gene that encodes this proteasome
49:10subunit beta beta 5T and the PSMB 5 is
49:15actually the the T stands for thymus,
49:19thymus specific. So I started with PSMB 5,
49:22but that's not the thymus specific one.
49:25So I found this paper and I found just
49:28that there were only four papers.
49:30That were published with immunohistochemistry
49:33with this diagnostic analyte.
49:36And I would just dig into the,
49:38dig into the methods and and you have to dig
49:42and then dig and dig it's detective work.
49:45So the paper that Anya was referring to
49:47that was published in AMJ Surge path,
49:49they said C prior,
49:50so I kept going back C prior, C prior,
49:53here's some supplementary material
49:55and it was a polyclonal that they
49:58had produced in their lab that wasn't
50:01commercialized and so was not available.
50:03But that's OK because I said.
50:06Basic science is $160 billion
50:10annual industry,
50:11and I said there are usually hundreds
50:14of commercially available antibodies
50:17to any immunohistochemical target.
50:20And so I just started this is bio compared,
50:23this is the dumb,
50:24this is the dumb database and
50:26this this antibody that Anya,
50:29who is an expert,
50:30she is the IHC lab director at Mayo,
50:33said this is an antibody that is
50:35not commercially available.
50:37There are 63 different commercially
50:40available antibodies to PSMB 11.
50:44Now that's a problem because I can't test 63,
50:47I can only test a few.
50:51So then I applied it to let's use a,
50:53let's use a finer screen.
50:55And then so here's the antibody pedia.
50:58And. And so again,
51:00we're looking for the green circles and
51:02so we've got some more promising targets.
51:05And then here's the bench side.
51:07And I ended up going with the bench side.
51:10And I found from looking at these figures,
51:13actually this figure here,
51:15this is a figure that demonstrates
51:18that I want to see which is nuclear.
51:21The nuclear staining in
51:23thymic epithelial cells.
51:25And so this is the antibody.
51:26This is the antibody that I that
51:28I proposed to Anya that she get
51:30and and work up in her lab for
51:32this diagnostic application.
51:34Now the icing on the cake is if in
51:36the last couple of months we actually
51:38had had the opportunity to do it.
51:40I pushed and pushed to do it
51:42for the uscap deadline and she,
51:44because she's European,
51:46was thoughtfully like on vacation
51:48for six weeks in Germany. So.
51:52And then actually this is,
51:54you know, this is a really,
51:55really good place to stop because
51:58here's the next step I said you can you
52:01can do all of the validation in silico,
52:04in silico.
52:05And so this is I guess who in the
52:10room knows what C bio portal is.
52:14Great.
52:15So everybody in the room should
52:18know what C bioportal is.
52:21It is this for me,
52:23it is the single most useful and
52:27user-friendly bioinformatics tool
52:30that you can use to analyze quickly
52:35and so easily both molecular genetic
52:38data and gene expression profiling
52:41data and the main source of the data.
52:46Is the TCG a date data sets and so
52:49there are pan cancer datasets and
52:52so you you go to the you go to C
52:56bioportal and you you can click like
52:58click a button and it says I want to
53:01analyze all the TCG a tumor types
53:03and it's like 30 tumor types and
53:05then at the bottom you just type in the gene.
53:08So I typed in PS:,
53:10MB 11 and then you hit a couple buttons like
53:13it changes a little bit from time to time.
53:16You have to know where to click.
53:17You click plots. And then I sorted
53:20these plots by median gene expression.
53:24And so in the TCG a data set.
53:28This is expression of PSMB 11.
53:31And it's nothing. Nothing.
53:33Nothing. Nothing. Nothing.
53:34Nothing. Oh my God. And guess what?
53:37In the TCG a data set.
53:39What does THYM mean?
53:42It means thymoma.
53:43So before Anya even does this experiment.
53:47I,
53:47I know that this is a a winner and
53:49so a lot of the the diagnostic
53:51markers that that I you know that I
53:54ran out of time to share with you.
53:56I do,
53:57I do this and I and I do this
53:59not only to sort of confirm
54:01specificity but also to identify
54:03additional diagnostic applications.
54:06And so the the best example of that
54:08is IGF 2 where it was published
54:11in this adrenal cortical carcinoma
54:13space, adrenal cortical carcinoma
54:15versus adenoma and Jason?
54:17My good friend said and I'll sort of
54:19stop with this story you said what do
54:21you know about this is this is how
54:23he interacts with me what do you know
54:25about IGF two and I said Ohh Sylvia
54:27and Oscar published it there and you
54:29know they're endocrine pathologist.
54:31It works great for this diagnostic
54:33application that I shared.
54:34He said I brought this antibody
54:36up and I'd love it and I want to
54:38do a project with it and we'll
54:40we'll what project can we do?
54:43And and I and I I said well we
54:45can't do the same project again.
54:47And so I said, well, let's take a look.
54:49Let's take a look at this in.
54:54I'll just pull up the because I want to
54:56show off see bio portal one more time.
55:01And these are all the different,
55:03you know, I I I use other tools to
55:06visualize gene expression profiling
55:07as well this GTX alumina and bio GPS.
55:11I often just use Wikipedia.
55:14But here's the C bio portal.
55:16I clicked tcga pan Cancer Atlas
55:19studies query by Gene, entered IGF 2.
55:25And here's the plot.
55:27And actually, so this is high.
55:32And this is high.
55:34And this is adrenal cortical carcinoma.
55:38And this is paraganglioma pheochromocytoma.
55:41And so by visualizing this,
55:44I knew that IGF two would be highly
55:47expressed in paraganglioma pheochromocytoma.
55:50And then I just have to go do
55:52the experiment to prove it.
55:53What's not in the TCG a data set
55:56next neuroendocrine tumor isn't.
55:58So I didn't know that.
56:00I didn't know if this meant
56:02that IGF two was going to be a
56:05general neuroendocrine marker or
56:06perhaps it would be specific.
56:08In the peripheral versus net differential
56:11and it was the IT was the latter
56:14that that I confirmed and IGF two.
56:17We talked about analytic robustness
56:19this morning with God at three and
56:22paraganglioma pheochromocytoma and
56:23I said TH is incredibly analytically
56:26robust but it's only expressed
56:28by sympathetic paragangliomas.
56:32IGF two is incredibly analytically
56:34robust that means it's expressed
56:37even if the sample set on the bench.
56:39For a month and it's nearly 100%
56:42sensitive in Paraganglioma pheochromocytoma.
56:45So again, like simple, simple tools,
56:50C bio portal, Oh my God,
56:52you know these databases. Me.
56:54You know, talk, talk to me, workshop,
56:57your workshop, your project.
56:59There are hundreds of commercially
57:01available antibodies to every analyte
57:03that you might be interested in.
57:05You can be a wildly successful
57:08anatomic pathologist.
57:09Even if your research budget
57:12is $300.00 a year,
57:13and I'll stop and take your and take your
57:16questions, thank thank you very much.
57:18I had a I had a blast.
57:19I always have a blast.
57:20I always have a blast.
57:29Questions.
57:32Maybe I start first.
57:35Vision of publications.
57:39One
57:40is that this gene so good
57:42for identified this protein.
57:45I got all the states, yeah, never,
57:48almost never to the original study.
57:53All my sins are there. Yeah.
57:54So in this situation, OK,
57:57so the basic search research
57:59on my thing doesn't work.
58:01So, you know, I I I will say
58:04because I'm very skeptical.
58:05Very skeptical. There are.
58:09I have had multiple instances where
58:12a new biomarker is published in
58:15the diagnostic pathology space,
58:18usually in modern pathology.
58:19And the one that comes to mind,
58:22there was a one.
58:24Like a wonderful appearing
58:26immunohistochemical marker that
58:28was published as the Diagnostic
58:30marker of anaplastic thyroid cancer.
58:33And they discovered it by gene
58:35expression profiling and and I got the
58:37antibody and it didn't and it didn't work.
58:40So so many antibodies don't
58:42work and that's the problem.
58:44And that's the problem.
58:46The problem many antibodies don't
58:48work because foreign antibody to work
58:51in immunohistochemistry is a mirror.
58:53Is a miracle because those antigens
58:56are so so deranged because of formal
58:59and fixation and because of antigenic
59:02degradation with cold ischemia time and.
59:05And so I was disappointed,
59:07like.
59:07I'm skeptical.
59:08I think that some science is that
59:11one of the reasons that it doesn't
59:13work is the science is fate, right?
59:16The the immunohistochemical
59:18results are are fate.
59:20Another reason is you know.
59:23That you know,
59:24it's not necessarily that the
59:26analyte is wrong.
59:27It's maybe that's not the best antibody.
59:30And so, like in that setting,
59:32I would encourage you to not give up.
59:36I would encourage you to at least
59:38try a second and a third time.
59:40And I would encourage you to
59:42try the antibody pedia and the
59:44bench and the bench side.
59:46I trust that more than a single
59:50published paper where the authors are.
59:53Uh, conflicted.
59:54They're incentivized to
59:55have a positive result.
59:57So.
59:58So I trust those databases
01:00:00more than any single paper.
01:00:04And then there are,
01:00:05you know,
01:00:05there are vendors that I intrinsically
01:00:07trust more because they do better
01:00:10science and like 2 great vendors
01:00:12for primary antibodies, abcam,
01:00:14wonderful cell signaling, another standout.
01:00:19There are many other, you know,
01:00:21again like exceptional, lovely, right.
01:00:23And these are exceptional.
01:00:25And then I'll mention one other
01:00:28thing that that you can try that
01:00:31you can try that I can't try,
01:00:33but I mentioned it.
01:00:34Is Don Pot asked about albumin ish
01:00:37this morning and albumin ish is only
01:00:41automate automated on a like a bond three.
01:00:43I don't have a like a bond 3.
01:00:46If you have a like of bond three
01:00:50you can make an RNA scope probe
01:00:54to any to literally anything and
01:00:57so and I've been workshopping
01:00:59this idea with a guy called
01:01:02Greg Sharvil who's a sarcoma.
01:01:04Pathologists at Stanford and and
01:01:07there are some, there are some,
01:01:09even though the genetics are on point
01:01:12and the gene expression is on point,
01:01:14that will never succeed as
01:01:17immunohistochemical assays because M,
01:01:19RNA and protein are correlated.
01:01:21But the R-squared is like
01:01:24not .9 and so there are some.
01:01:29You know,
01:01:30some targets that are incredibly specific,
01:01:32but there's such low abundance
01:01:35that you can never,
01:01:36even though we do signal amplification
01:01:39and immunohistochemistry,
01:01:40you can never get the
01:01:42immunohistochemistry to work.
01:01:44And for those analytes,
01:01:46if you really are committed to a,
01:01:49to a, to a specific target for a
01:01:52specific diagnostic application,
01:01:54I would encourage you to, you know,
01:01:57first try a couple more immunos and try.
01:01:59Try bench sign antibody pedia.
01:02:01But but the next thing I would
01:02:03encourage you to consider is
01:02:05building an RNA scope assay to that,
01:02:08to that, to that, to that target.
01:02:10Because if the you know the M RNA don't lie,
01:02:13the M RNA might not correlate with the
01:02:15protein, but the M RNA don't lie so.
01:02:19What's your quick?
01:02:22The communist.
01:02:25Sure.
01:02:28Sure.
01:02:31Yeah.
01:02:361.
01:02:39Sure. Yeah, I mean I, you, you,
01:02:42you, you probably would not be
01:02:44surprised that I give like.
01:02:461st 1st it's really really hard
01:02:48for me to give a 45 minute talk.
01:02:50I usually I usually speak for two
01:02:53or three hours at a time and and of
01:02:56course I have hour and two hour and
01:02:593 hour PDL 1 lectures and and I I
01:03:03mentioned you know immuno optimist,
01:03:04immuno pessimist,
01:03:05immuno realist and you know I give a
01:03:10PDL 1 lecture that is called PDL 1
01:03:15immunohistochemistry like you know.
01:03:17The Emperor has no clothes so uh
01:03:21Pete and I I gave I gave a biomarker
01:03:24lecture in a global in the global
01:03:27health space this summer and there
01:03:29were lots of questions about PDL
01:03:32one and and and there's a YouTube
01:03:34clip of me saying something like
01:03:37PDL 1 immunohistochemistry sucks.
01:03:39I hate it and it just loops and loops
01:03:42and loops and so and so my my main take on.
01:03:47PDL One immunohistochemistry is.
01:03:50It's PDL one is super important.
01:03:55It was a great, it was a great target,
01:03:59like Nobel Prize, like totally warranted.
01:04:02But it is an entirely unrealistic
01:04:07expectation for, in that space,
01:04:10a single analyte to reflect the
01:04:13entire tumor immune microenvironment.
01:04:17So PD1 immunohistochemistry is
01:04:20an inaccurate diagnostic test.
01:04:23It's about 60% sensitive, 60% specific.
01:04:27It's a coin flip.
01:04:29And they're ohh I I really, really,
01:04:35I really, really hate capitalism.
01:04:39And uh, Pharma really,
01:04:42really loves capitalism.
01:04:44And so, you know,
01:04:47the the drug companies,
01:04:49we're going to partner with
01:04:51the device companies and jam a
01:04:53companion diagnostic marker down
01:04:55our throats and they don't really
01:04:57care about the diagnostic accuracy.
01:04:59So PD1 immunohistochemistry sucks.
01:05:01I hate, I hate it.
01:05:03That being said, I spend,
01:05:05you know,
01:05:06hours and hours of my life talking about it.
01:05:09And I read every PDL 1 immunohistochemistry
01:05:12at the University of Iowa and
01:05:15PDL 1 immunohistochemistry.
01:05:16You know this is a good example
01:05:18of a bad of a bad marker.
01:05:20There's biochemistry problems
01:05:22and there's readout problems.
01:05:24So like pathologist interpretation
01:05:26problems and they're and they're unsolved.
01:05:29And my like skepticism around
01:05:32capitalism is pharma did not
01:05:34consider the biochemistry or the
01:05:37readout in developing the.
01:05:40The this diagnostic marker because
01:05:42they actually have no technical
01:05:45expertise in immunohistochemistry.
01:05:47So what's better than PDL one not TMB.
01:05:51TMB is just as bad.
01:05:53About 6060 MMR is wonderful
01:05:56if you have an MRI,
01:05:58MRI,
01:05:59immunohistochemistry or MSI is very accurate.
01:06:03Not 100%,
01:06:04but very very very you know high
01:06:08ninety 9899% accurate and.
01:06:09Is very predictive,
01:06:11not 100% percent predictive,
01:06:12but very predictive of response
01:06:14to checkpoint inhibition.
01:06:15And then I've learned so much from
01:06:18my collaboration and interaction
01:06:20with David over the last five years.
01:06:23The answer to the answer to the
01:06:28checkpoint inhibitor predictive marker
01:06:30is Multiplex immunofluorescence act.
01:06:33Actually I said it's unrealistic for
01:06:35a single analyte to integrate the
01:06:38entire tumor immune microenvironment.
01:06:40And there's a lovely meta analysis
01:06:44that was published by Janis Tobb,
01:06:47who's one of David's collaborators.
01:06:48She's a Melanoma doctorate at Hopkins
01:06:51that showed that the area under the
01:06:55curve for Multiplex immunofluorescence
01:06:58for this diagnostic application is
01:07:01vastly superior to immunohistochemistry.
01:07:04And then the problem is that's a huge
01:07:07paradigm shift, like there is no training.
01:07:11In Immunohistochemistry lab directorship,
01:07:13I taught myself. I taught myself.
01:07:16I try, you know, I really,
01:07:18really care to try to teach you.
01:07:21And then to try to teach people to
01:07:23bring up Multiplex immunofluorescence
01:07:25for this diagnostic application.
01:07:27It's a really, really heavy lift.
01:07:30But possibly, you know.
01:07:34And this is perhaps controversial, possibly.
01:07:37And my thoughts on this have
01:07:39changed quite a bit over time.
01:07:41Possibly every lab test should not
01:07:44be offered by every lab and so,
01:07:47so maybe the solution is Multiplex
01:07:51immunofluorescence in 10 or
01:07:5320 labs that have the the,
01:07:55you know,
01:07:56the resources and the expertise to deploy it.
01:07:58So thank you for asking me about
01:08:01my least favorite stain.
01:08:17Yeah, I can and so I can, I can,
01:08:20I can talk for half an hour about TRPS,
01:08:221 TRPS. One is classic Gartner hype cycle.
01:08:26The study that was the study that
01:08:28was designed was highly flawed.
01:08:30Highly flawed, yeah.
01:08:31And that and that's on the.
01:08:34Uh, that's on the authors. You know,
01:08:37they should have designed the right study.
01:08:39And it's also on the reviewers
01:08:40for not identifying the flaw.
01:08:42Like, briefly, TRP's like,
01:08:43do you want me to talk about TRPS one?
01:08:46Yeah.
01:08:51Yeah.
01:08:56That's it.
01:09:01Yeah.
01:09:05Yeah, sure.
01:09:11This model and yeah.
01:09:17Ohh yeah, yeah. So. So this is the,
01:09:21I mean I said this is the in silico
01:09:25validation and then the validation
01:09:27is I always do it's this way is
01:09:31actual validation and then I'll say
01:09:34I say floor and then I do much more
01:09:36and then I'll briefly just I'll
01:09:39because it's a good example I'll
01:09:42briefly describe how I validated TRPS
01:09:44one and actually if you want to.
01:09:47Know how I validated TRPS 1I tweeted
01:09:50how I validated TRP PS1 because I
01:09:53because I I thought the my findings
01:09:56were sure were significant and so the
01:09:58floor to to validate a diagnostic
01:10:01marker is 10 expected positives
01:10:03and 10 expected negatives.
01:10:05And I almost never do the floor
01:10:08because if you do the floor then
01:10:10there's all kinds of diagnostic
01:10:13pitfalls that you will never identify.
01:10:16I mainly identify these.
01:10:17Potential diagnostic pitfalls by looking
01:10:19at this gene expression profiling data,
01:10:22mainly using C bioportal.
01:10:24So for my validations I use tissue
01:10:28microarray almost you know a combination
01:10:30of whole section and tissue microarray.
01:10:32And for TRPS 1I validated TRPS
01:10:36one in at least 100 breast cancers
01:10:41and at least 100 non breast cancer
01:10:45non breast cancers.
01:10:47Mainly carcinomas,
01:10:48but other tumor types as well.
01:10:51Like it's imperative to validate
01:10:54a new diagnostic marker in a fit
01:10:58for purpose fashion and that means
01:11:01in consideration of differential
01:11:04diagnostic considerations.
01:11:05And I'll tell you what happened
01:11:07with the with the,
01:11:08a couple things with the TRPS one
01:11:12and then another thing that I talk
01:11:14about when I'm doing my straight
01:11:16didactic about immunohistochemistry.
01:11:18Talk about added value of semi
01:11:21quantitative IHC stain assessment.
01:11:23Positive and negative is not enough,
01:11:25you need to at least semi quantitate
01:11:29the signal. In the tier.
01:11:31So when I went to see Bioportal
01:11:33actually I don't know can we go to,
01:11:35can we go to,
01:11:36let's go to see bio portal can
01:11:38I can I back out of here.
01:11:39So
01:11:40maybe you can discuss it later by
01:11:41yourself. Yeah, you're going to have
01:11:43do we have more questions?
01:11:46Let's hit another question.
01:11:49Ohh no, I'm screwing everything up.
01:11:52I think you know the question.
01:11:53Later. Not sure
01:11:56PRPS one, we'll talk about all.
01:11:59We'll talk about all afternoon.
01:12:00Thank you so much Andrew.
01:12:02I think I'm going to eat.