From 628dc2713926b339d683826c8afc991d0cd62432 Mon Sep 17 00:00:00 2001 From: January Weiner Date: Thu, 3 Oct 2024 17:53:07 +0200 Subject: [PATCH] updated presentation --- how-to-talk-to-your-bioinformatician.html | 267 ++++++---------- how-to-talk-to-your-bioinformatician.qmd | 360 +++++++++++++--------- images/pregnant.jpg | Bin 0 -> 28004 bytes 3 files changed, 315 insertions(+), 312 deletions(-) create mode 100644 images/pregnant.jpg diff --git a/how-to-talk-to-your-bioinformatician.html b/how-to-talk-to-your-bioinformatician.html index 81ba9b8..2834ee4 100644 --- a/how-to-talk-to-your-bioinformatician.html +++ b/how-to-talk-to-your-bioinformatician.html @@ -3134,64 +3134,22 @@

How to talk to your bioinformatician?

-
-
-

Part I

- +
+

Who am I to tell you things?

-
-

Slide 1

- -
-
-

Bioinformatics vs statistics vs computational biology

+
+

Bioinformatics, statistics, computational biology

Things bioinformaticians care about

    -
  • Experimental design
  • The biological question
  • Statistics
    • +
  • Experimental design
  • Reproducibility
  • File formats
-
-
-
-

Experimental design

-
-
-

How many samples are sufficient?

-
    -
  • Depends on the question
    • -
  • Depends on the technology
    • -
  • Depends on the variability
    • -
-
-
-

How many samples are sufficient?

-
-
-

The bottom line

-

Talk to your bioinformatician early!

-

Statistics

@@ -3293,6 +3251,80 @@

Why is that important?

+ +
+
+
+

Experimental design

+ +
+
+

How many samples are sufficient?

+
+
+
    +
  • Depends on the question
    • +
  • Depends on the technology
    • +
  • Depends on the variability
    • +
+
+ +
+

+
+
+
+

How many samples are sufficient?

+

Say, we want to compare two groups with a standard \(t-test\), nothing fancy. Our ability to detect the differences (the statistical power) depends on the sample size and the effect size1.

+
+
+
+
+

+
+
+
+
+
+
+

How many samples are sufficient?

+

The \(y\) axis on this plot shows how the power of the test – meaning how often, assuming that the groups really differ by \(d\) on average, you will be able to detect the difference using a t-test.

+ +
+
+

How many samples are sufficient?

+

What about the following setup:

+
    +
  • We have 2 strains (WT and KO)
    • +
  • We have treatment + control
    • +
  • We want to know whether the treatment has a different effect on the KO strain than on the WT strain
    • +
+

This is a 2x2 design, and we need to consider the interaction term.

+
+
+

How many samples are sufficient?

+ +
+
+

How many samples are sufficient?

+

That is not even the worse thing.

+

Simple calculations show that assuming

+
    +
  • your power is 80% (really great!)
    • +
  • \(p-value\) cutoff is \(0.05\)
    • +
  • 90% of the \(H_0\) are true (i.e., 10% of the time the differences are real)
    • +
+

then 36% of your “significant” results are false positives1!

+

(Plus, you failed to detect 20% of the real differences)

+ +
+
+

The bottom line

+

Talk to your bioinformatician early!

@@ -3422,6 +3454,26 @@

How we work

  • Fine-tuning the analysis results
    • +
    +

    Identifiers

    + +

    Excel and gene names

    @@ -3436,7 +3488,7 @@

    Excel and gene names

    -
    +

    Excel and gene names

    @@ -3513,7 +3565,7 @@

    Don’t combine values and comments

    -
    +

    How (not to) work with Excel

    @@ -3587,11 +3639,7 @@

    Thank you

    You can find this presentation along its source code at https://github.com/bihealth/howtotalk

    -

    Sources

    -
      -
    • Source 1
      • -
    • Source 2
      • -
    +
    @@ -3603,117 +3651,6 @@

    Thank you

    -
    -
    -

    Slide 1

    -

    To compile, type quarto render template.qmd

    -

    Make sure you have Quarto 1.2 installed from here.

    - -
    -
    -

    Multicolumn slide

    -
    -
    - -

    Left column title

    -

    Left column…

    -
    - -

    Right column title

    -

    Right column (60%)…

    -

    (adding .fragment causes the contents to be displayed in steps)

    -
    -
    -
    -
    -

    Part II separator slide

    - -
    -
    -

    Simple numbered and unnumbered lists

    -
    -
    -
      -
    • One
      • -
    • Two
      • -
    -
    -
      -
    1. One
      2. -
    2. Two
      3. -
    -
    -
    -
    -

    Incremental list

    -
    -
      -
    • Item 1
      • -
    • Item 2
      • -
    • Item 3
      • -
    -
    -
    -
    -

    Incremental contents

    -

    First part

    -
    -

    Second part

    -
    -
    -
    - -

    This is a slide without a title (use the dashes to separate)

    -
    -
    -

    Transitions

    -

    Define them in the YAML header or like here, in the slide title.

    -

    Types: none, fade, slide, convex, concave, zoom

    -
    -
    -

    Code

    -
    -
    -
    plot(1:10)
    -
    -
    -
    -
    -
    - -
    -
    -Figure 1: A dumb plot -
    -
    -
    -
    -
    -
    -
    -
    -
    - -
    -

    Tip

    -
    -
    -

    Ctrl-click on the image to zoom. And here is a 3.1415927 for you.

    -
    -
    -
    -
    -
    -

    Code

    -

    There are many customization options for the code. For example, you can highlight (and even animate) certain lines of code:

    -
    -
    a <- rnorm(10)
-b <- rnorm(10) + a
-c <- a + b * rnorm(10)
    -
    -

    You can also specify where precisely should the output of the code go: below the code (default), on the next slide, on a right-hand column…

    diff --git a/how-to-talk-to-your-bioinformatician.qmd b/how-to-talk-to-your-bioinformatician.qmd index 814bad6..3f27912 100644 --- a/how-to-talk-to-your-bioinformatician.qmd +++ b/how-to-talk-to-your-bioinformatician.qmd @@ -27,41 +27,18 @@ format: --- -# Part I {.inverse background-color="#70ADC1"} +## Who am I to tell you things? -## Slide 1 - -::: {.notes} - * Bioinformatics is not a service - * Bioinformatics is not IT - * Bioinformatics is not a black box -::: - -## Bioinformatics vs statistics vs computational biology +## Bioinformatics, statistics, computational biology ## Things bioinformaticians care about - * Experimental design * The biological question * Statistics + * Experimental design * Reproducibility * File formats -# Experimental design - -## How many samples are sufficient? - - * Depends on the question - * Depends on the technology - * Depends on the variability - -## How many samples are sufficient? - - -## The bottom line - -### Talk to your bioinformatician early! - # Statistics ## Statistics matters @@ -120,6 +97,208 @@ You have to understand p-values and their limits to talk to other scientists! ::: +::: {.aside} +Wasserstein RL, Lazar NA. The ASA statement on p-values: context, process, and purpose. The American Statistician. 2016 Apr 2;70(2):129-33. +::: + + +# Experimental design + +## How many samples are sufficient? + +:::: {.columns} +::: {.column width="45%"} + + * Depends on the question + * Depends on the technology + * Depends on the variability + +::: +::: {.column width="10%"} +::: +::: {.column width="45%"} + +![](images/pregnant.jpg) + +::: +:::: + +## How many samples are sufficient? + +Say, we want to compare two groups with a standard $t-test$, nothing fancy. +Our ability to detect the differences (the statistical *power*) depends on +the sample size and the effect size[^cohen]. + + +```{r echo=FALSE} +#| label: effect_sizes +#| fig-width: 9 +#| fig-height: 4 +# plot four effect sizes as boxplots +library(tidyverse) +library(ggplot2) +library(ggbeeswarm) + +n_samp <- c(3, 4, 5, 7, 10, 15, 25) +d <- c(Medium=0.5, Large=.8, "Very large"=1.2, Huge=2.0) +n_samples <- 100 + +df <- imap_dfr(d, ~ { + d_val <- .x + size <- .y + + x <- rnorm(n_samples, mean = 0, sd = 1) + y <- rnorm(n_samples, mean = d_val, sd = 1) + + data.frame(value = c(x, y), group = rep(c("A", "B"), each = n_samples), d = d_val, size=size) +}) %>% + mutate(size = paste0(size, " (d=", d, ")")) %>% + mutate(size = factor(size, levels=c("Medium (d=0.5)", "Large (d=0.8)", "Very large (d=1.2)", "Huge (d=2)"))) + + +ggplot(df, aes(x = group, y = value, fill = group)) + + geom_boxplot(outlier.shape = NA) + + geom_beeswarm(alpha=.25, corral.width=1.5) + + facet_wrap(~ size, ncol=4) + + labs(x = "Group", y = "Value", fill = "Group") + + theme(legend.position = "none") +``` + + + + + +[^cohen]: Here we use Cohen's $d$ + +## How many samples are sufficient? + +The $y$ axis on this plot shows how the power of the test – meaning how +often, assuming that the groups really differ by $d$ on average, you will +be able to detect the difference using a t-test. + +```{r echo=FALSE,cache=TRUE} +#| label: power +# we could use pwr package here, but simulation is more explicit + +# parameters +n <- 1000 + +# simulation +power <- map_dfr(n_samp, ~ { + n_samples <- .x + map_dfr(d, ~ { + d_val <- .x + pwr <- map_dbl(1:n, ~ { + x <- rnorm(n_samples, mean = 0, sd = 1) + y <- rnorm(n_samples, mean = d_val, sd = 1) + t.test(x, y)$p.value + }) < 0.05 + data.frame(n_samples = n_samples, d = d_val, power = mean(pwr)) + }) + +}) %>% + mutate(study = "Simple t-test") +``` + +```{r} +#| label: power_plot_simple +#| fig-width: 9 +#| fig-height: 4 + +ggplot(power, aes(x = n_samples, y = power, color = factor(d))) + + geom_point() + + geom_line() + + scale_color_discrete(name = "Effect size") + + scale_x_log10() + + labs(x = "Number of samples", y = "Power") + +``` + + +## How many samples are sufficient? + +What about the following setup: + + * We have 2 strains (WT and KO) + * We have treatment + control + * We want to know whether the treatment has a different effect on the KO + strain than on the WT strain + +This is a 2x2 design, and we need to consider the interaction term. + +## How many samples are sufficient? + +```{r echo=FALSE,cache=TRUE} +#| label: power_int +# we could use pwr package here, but simulation is more explicit +library(broom) + +# parameters +n <- 1000 + +# simulation +power_int <- map_dfr(n_samp, ~ { + n_samples <- .x + map_dfr(d, ~ { + d_val <- .x + pwr <- map_dbl(1:n, ~ { + x <- rnorm(n_samples, mean = 0, sd = 1) + y <- rnorm(n_samples, mean = 0, sd = 1) + z <- rnorm(n_samples, mean = 0, sd = 1) + w <- rnorm(n_samples, mean = 1 * d_val, sd = 1) + df <- data.frame(x=c(x, y, z, w), + strain=rep(c("WT", "KO"), each=2 * n_samples), + group=rep(c("Control", "Treatment"), each=n_samples)) + mod <- lm(x ~ strain * group, data = df) + anova(mod)$`Pr(>F)`[3] + }) < 0.05 + data.frame(n_samples = n_samples, d = d_val, power = mean(pwr)) + }) +}) %>% + mutate(study = "Interaction") +``` + + + +```{r echo=FALSE} +#| label: power_plot + +power_tot <- rbind(power, power_int) %>% + mutate(study = factor(study, levels=c("Simple t-test", "Interaction"))) + +ggplot(power_tot, aes(x = n_samples, y = power, color = factor(d))) + + geom_point() + + geom_line() + + scale_color_discrete(name = "Effect size") + + labs(x = "Number of samples", y = "Power") + + facet_wrap(~ study) +``` + +## How many samples are sufficient? + +That is not even the worse thing. + +Simple calculations show that assuming + + * your power is 80% (really great!) + * $p-value$ cutoff is $0.05$ + * 90% of the $H_0$ are true (i.e., 10% of the time the differences are + real) + +then 36% of your "significant" results **are false positives**[^more]! + +(Plus, you failed to detect 20% of the real differences) + +[^more]: I can walk you through a very simple demonstration later if you +care. + +::: {.aside} +Colquhoun D. An investigation of the false discovery rate and the misinterpretation of p-values. Royal Society open science. 2014 Nov 19;1(3):140216. +::: + +## The bottom line + +### Talk to your bioinformatician early! # Reproducibility @@ -237,6 +416,14 @@ In the diagram above, two things take usually the most hands-on time: * Data cleanup * Fine-tuning the analysis results +## Identifiers + + +::: {.notes} + * Identifiers are unique + * Identifiers are stable + * Identifiers are machine-readable +::: ## Excel and gene names @@ -461,10 +648,6 @@ at [https://github.com/bihealth/howtotalk](https://github.com/bihealth/howtotalk ::: ::: {.column width="40%"} -Sources - - - Source 1 - - Source 2 ::: ::: {.column width="20%"} @@ -479,122 +662,5 @@ plot(qr_code("https://github.com/bihealth/howtotalk")) ::: - - -:::: - -## Slide 1 - -To compile, type `quarto render template.qmd` - -Make sure you have Quarto 1.2 installed from -[here](https://quarto.org/docs/download/prerelease.html). - -::: {.aside} -This aside note can be used to add citations or notes -::: - -## Multicolumn slide - -:::: {.columns} - -::: {.column width="40%" .fragment} -### Left column title - -Left column... -::: - -::: {.column width="60%" .fragment} -### Right column title - -Right column (60%)... - -(adding `.fragment` causes the contents to be displayed in steps) -::: - -:::: - - - -# Part II separator slide {.inverse background-color="#70ADC1"} - -## Simple numbered and unnumbered lists - -:::: {.columns} - -::: {.column width="50%"} - * One - * Two -::: - -::: {.column width="50%"} - 1. One - 2. Two -::: - :::: -## Incremental list - -::: {.incremental} - - Item 1 - - Item 2 - - Item 3 -::: - - -## Incremental contents - -First part - -. . . - -Second part - - ---- - -This is a slide without a title (use the dashes to separate) - -## Transitions {transition="zoom"} - -Define them in the YAML header or like here, in the slide title. - -Types: none, fade, slide, convex, concave, zoom - - -## Code - - -```{r} -#| echo: true -#| output-location: column-fragment -#| label: fig-sample -#| fig-cap: A dumb plot -plot(1:10) -``` - -::: {.callout-tip} -Ctrl-click on the image to zoom. And here is a `r pi` for you. -::: - -## Code - -There are many customization options for the code. For example, you can -highlight (and even animate) certain lines of code: - - -```{r} -#| code-line-numbers: "|1|2|3|" -#| echo: true -a <- rnorm(10) -b <- rnorm(10) + a -c <- a + b * rnorm(10) -``` - -You can also specify where precisely should the output of the code go: -below the code (default), on the next slide, on a right-hand column... - - - - diff --git a/images/pregnant.jpg b/images/pregnant.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e3e8283c41f1b399865f68d5441e7e1780fa8f80 GIT binary patch literal 28004 zcmb5Vb9f~|_b$4lnb@{%XJTVVlVoDsw(VqM+qN;WZQFJ-@yUGO?>zV1`<#DncR#JQ zdey3`cduUkR_)K_&usvrl$fL#00aaCAn~;VKGy)k0LX9OzJG&&{Qmtr6ci*hEGir< z3=Aw5G717J9yS3!9yTs6At^ltAu%lpE-obp6)htZD=RAj1s5+TGY>rrEAu}iAW%?H zurRQgaB!H+M7TuE|G$^dUH}p#h%?AN7zi-{6bS?j3FNaMfCB&lfP#H_`~NK95TIb+ zK)!zovA*vAzhwXrP%v1{I69w52`tO6ApZn z;*FuJABmc*j_@`@?lx{-Lo!F)fHMxobrI127=?ijYZf`={n1QA+ohRAE3^*DNJL3M}*8igvR0$=R~fC3Z2(IHdJ z%qiW_B@wMht@+_4Sv@#xI>!e6>iUts3b{GW(ExYF3jzR96n6u4wpmM<>}WC!MadhL zRWCA2$^T6ZS4!gvVWn3^&Vuq5RnO}pTTz?voBi3{EFE2n0`?9Z5by)7@(*1G*J>5% z>dTptiOlrmFPSU~SZCeTsGGObv-mr{copRxa^aG#Aye)U2I_dvJ)%wSj=%7?^<}b^ znOKHw;{ZZsjU^2;?dLiQ+?2GACr$MsxEA#WM=vebek$aWa!cKVV%~k}s=Bllb$Pi) zcQEdwVrLxk`fqOj0D){Ri&e4M!b&HjPEnVwnL9k51D$ABvT9?*dVDJjJA-`6i+JQ~ zzfNxF(+DJb@}E^TH6eNlCu`+4vTi&!F@Tg>qh~K74}z4cOm9v2`KPt#%Q(|J1|)p6 zNYbU-jcTl{8TMhlxjf~{S zh!$%|wW9-_w?tTC0`7D9*y$<7fE%1vfoXQerpe`PS>t1c!F@YKkYUH!<%NgKFoOUf zO3vX9c!r=nUf$d)F_m2`E;;pV{?1LWQnr>1shBh2XwQN6*Ku?%9zTWCCdzb@?}~NM zDv-v*#34fs=WYUP9{|acyLo3n#GZJc-1uwt&nFzuhK;l_{+Qsl^^~s2z?q=ii{0EKG zRg{L&yBJ{cdik=$T)4?f`+=|uVyVCSIZ$frVN%GLU69w zYBz4Si}dcJ7kc(|f!`QDe18Y((}A~-FS!@V^UgxGbGUiqw-g8I)zofnyGtty7BZ`= z4`aNwO_>OD-fWM`fFl5a)jm@@-J%CL=0lNRBpc?WqM{{``nX8`T==L!_+ZyJW@vFl zo}Y8a+j8YRFjTw}OIHbOy(zRAE9UheuNQOk1rz(b*pQjlG#S&X-JDR>;r$6vwVDTR z2u44&csRM8?2B*B47HEfrp?)Y0;F>fPo6(eIu~nMZRX6+`iv`?YhM`#KBk9hLII$T z62FO;_K6%5-n^rW)K9nsGO^6B3^wi!X7`fYZ!Ko%=oIi1W~$3CwB>RSxGfaUCeo;i zTmLHSPB8+6Q{k6liD^qayfw-yqBymho^KyFL}D>;dbSj+R-MopkDT0IJ?Ugrufn1e z7C0JJo$WA9irYhm{!CDKr62K&MfNfQ03;^WG$$HB`sRq~Cv{#M8pw2V=CS>ks$UL>CZaE{(1<$R;5m@ zxu@bd-aG-r>Txd!q*L)Qr=AeFIrn#pu_uaFfwWf^WLH_$H|2j4pBxuAP*;AZop`0r ztgpt4u>!tDU0;|zn@_HqFH!um%0^h99@sOz!^b~%N_tpwWWO$RZ+WTKy9(V%!jfTL z%79_I<=s>l94#_BNgU#y(Lk)y1%N{;>R-6Fn1oNHRh;1LTe_v4-L{7<^2J_pboEuS;5@#vRnt>;^zG8JVu$cy8>k z(nWHmR6=?i$DLT#w3Wkbu>S%kx1o6yo+ccyHCbsciTAl2EsW-6HjJ#vIabVMf&hqQ z<}9qAcx9Y8EFuJgm@BP#gYZ!>Z*-J*PdF6I&S}gil-Dcw3X@d~xys+&P1I%%Ll?e* zx2<}%y+RGsqyRu*l-JoiyqNXI$3~DNt}>i!T5ZOpPjBiIs?Vt`BbsxKyzR5w8X8?E zw{#Lr5WF?|wKiW1%+|8w|5JB-5e;?gjtM6Z0w){LfV}X`#O7FHXSXEeD0VGrxrWt( z2_vnZbn|BlJO?$_b~j`fNi`}P>(&o{X6QozK&Wqvy{&DsPQmFf``KE!;|_%+u+Gp_ zEXWI(RmaVc?ZUGreK+-^weAkJZeSl#)-<&uvi&ZRhYPiwC`MQaQ|dAN&%1z z|F$r})-*j1Yc?KTMd6oF6KG&t7D}e9Irj`(Yj-X`kjXpjWVbhn{^#(m*FqDortfOAuDoMq>kJ{o@;SiJnfMCd1;bShmUM!+b>?2{b#3SrXj}$x) zxp`;>RH1KPN3lGc6Dy2hC?a{FoZ1P29D)qfZwHx#{quC?j%yOqhr3Tg6RRL+Q_S+CPo!Pq-$E%)p-tCb- zg{I8kldlu4)q}3Spx`eW+E;Q6aLsD_W_D~1nWCQRd1yntYT5Z-2{PcVeEnw|EK1|h zD>3#bV`7zxS;w0ORTy_}GIz2=mCpoA48AIi05I9G!kl7qLW*3ll@Y;5(~$Ez;%C~F ztWhM!48m7dp<6pfGAlx20LaKgtd%uCeaS3Ol;5GSa;vt;z>MC3aXHLy6vcx-DuwV_ z$-N;o4MWtY-?1<{2XGPQ2&ck>%>SD>{qH^K|72#Q#GnHI``-We_%HorLIi+-gMdPS zgMxy9f2EWF04M+i3>*mp84(4Qh>4Jxk(q==KoE_Tj8#}jUf1?3^Mw3LN|76`m3Q}h`;5T^^|4@{=~tA`dVfr;eJ@(2QSOsFvs7x- z-sq*tp#X}2SpAq7nY>@3x5~I2a7ncW7MHAev6o1Lyun#R`m({zlbl&*5;Ypbka+L! zz3R;ne^=+xU3|ca6E_DYlgwQJojYle*L|}#Qgnc)h7o_`=Af)j9}XLSWU?e-vM%ch z7K&o2`Bn7Erp&@%J1UO97|IqFA+OaI59lk*=1_O_G|SV&7vskLObcrYlar37@FJ(X zLg8V%@2)62$yKmW%5TaBM&{ItaU-e~TM(R4}{Px8h-b`lFcR=4| z_9M8)pNB)l9EKz8$ZC$kbaenZ_`ut;JLkXcu~*J8O5!wXB`zeb6HRBHLJO-x(#RYf z4!@B?L8q(vg7*3RQlS>2R8NpmJ94@IE;?V@h_l6eMAHIJcybvP8 zb5@e*q)?S-G%wqFN5q9^xF};Ct3qt|@I-@KtiTeUF-#-L6=fc*C_JitaH+ljx)Mxo zL_5<>W)2bxEA1-E>cCMDDX52dySCs~hxFIjC>m1%I^rzwyf~`f*QnDY)6%IKp+k|@ z<##YTnvGxAerbp7)l7W*>|8b@#j7m?w5gbD*&LS>)KSI*!7Vxg5XL_7RlwZ}2AzH} zIP$TrQr(Saw$gn16whjeTHDQ}9*0|;;nW5{oh|{+-&j?3+iDe`J*19+Sz;z^QZ`Fm zKmzKPJYQz8t7ofJ;ctwCD%Oh(H&20%HN;+^V_x_|IO7U?u2#8Vp>j1{+RSRWd8}%1 zC#D5jT8}w-A&Qz8c`4qPiU2pL&ip1X^Lu8ERUZx2NP>>xsoih`)Sl^}f{V*d;6@s| z+{&+XdLG@scwNkiomE`nEgx`PO%WttRl=p+3a$t!S^BY4{);_8p;}*f;S>Z|suEPb z=|8Sl{)>xzfkq|iumnk=*w9Kv;iKs0B>swqoP6-%%4yVbx=zn{*J>qI)v_Vza>^O? zJl|(1y89KDfUcK-ui0mORIa413=@yXIi8B|4W^Gz%{*GT42~NDWIE|H%|)8!}aqefNF(BRA?HgJWiOQKLWGgkthmX3rY&4lj)`Lk@lIMzM6iQL(F+bc_U z?>&Ge(sv9h8^&Q*%u@wfKK%~scKZ3@6R=L8aLr%GY+S3EUP%+vAEDqoZnF7W-}f?A zn{L5f&(BQLjU);qBO`q_bZ<$bqM| z=t(z~yPep@c~FFLthbs=Oz3Ud8;WmVHO6bq#PlZ^g;5Yz2lD{o1WGA)m@}2~jwlN= zRpus>3&rJWx<@`PSY7qC&njMv&p`?Ph{z{^v|!gNM7-62I%u|%3WZfK8Qs`OpWc#X zs7cI(a@_~(4+?){CQFf5wb%;C+xJ(x5c+lfCc4hY$9ULh%aIFc@E*wt6FL@;e&*uB z5a)Hmi8zzKX6TvoD2(t=0N`V0TieORxwQ*kLsC@7p1K=Wzjuv$p597e&y=5p8wRPq zzKsE{CsXAU5PywvzHp{5hy{@*NQB_g<^RnkbgZNfx|F4#4=wnoZp!ZZByKIa&f)>U zfh?*{V!_9k-@;A&NCkFOKgY6Dd29;vf)>e=+*RC?Xz&~>T+CVaa&jVWZKWNF1N!~^ z4dRtvc&dynm{8v|9cGMsr`?9ZL2(R+MfcAfM-zscDCQ^&_W+I2vK|#mm+W;k1b+%2 zh^JKlt_)Tg9w_*^Vuh{_3Tx&fhNRuG>GL z@8SF0Cf}BURIjzss7Vsv66VkbU3%%yd!k{h7Y}`fQ6J1!sP7grvOH<8L#C>+5;WkW z%>7ns?1a%*@NNeu8Vn(2fil!(S0m_G-5V%7#bJ?EV%@i1^=m=K1HXN{^#4qZS)cr! zSO%bHQB3XxHCkySa$jaeASd5ck^=MOmb*%LUT#v=YK~uNG@z0F8Ep?nNI065#aj!W zTD!UR@Hs~lW1$;%<#Am0WYYur!83o`w|xRm`rDD7Hk(O2^SoDWyC`7HhtRDKby_uk z-}U0Qxb___ZZ+FD5odIPcg}l=*?aOtcSi0ZNnI?9Pa#`2bS6~MLAlwTmnphUrHu1~ zcR7_v5Az@v@1V3|_F`d)rl(M4W`}X-;^k9Z8veXm80z&#e17}zm}<}wjN z>a~XT<58_p_-Ohk2uTk3;(b1vc3|+8Oh$ghT0Pt~@TuM;f&MiVrH)XR(no)Ey;)D3 z`N$=#Mdnov#-|M-8sU(Dujai2t}canoVE|&z{Ht7_R-g^{yf3NH5Uv+ZK_}*2)Er= zJZ-&5Lo{KMkN_gYSzKviGq>*G>kFcp*|RO09AkZqwwZZ-Bm%m{K`tn3lC zr|FR_VB6~L6JS3D<_Y6k);qvTrXZQ^TBE0K8KoMmMcomE>#ruQ9UY$%bbIp1^msl7S0bk%% zo5Gc7&OIFf81@j2Ap0ti;x>%KiN?9OyHW*%D(~B=?ke*TVPo56);#E2_X_TfR9o9+ z#VL|X&E(`F|K39E`kn(tOjHUBarO203!PQ+$7J^huf02@eE;RI62zdT;Efr4E%QWqHq<5R+U)7jkl zMgw^u<|ao$FR?$i&Qp^8PSIm9Mj`%)%BNVgLunKeoQtlX|D#GR%E)Byd$#Jw{;5uV zd|h~-Lor?*lmkiZwd2^PWgh+hD}Dh3Yu2xTc<%iv#?;p6%#ms6Z_CVj2&Q~6keo6H zG0CmW*I_M~7j_E4e_sdaXkZ2I>W-?s6(zGp2ml`&7pZV}4RcJo`H`{QV-RkP4!_$< zc6H89AC25le{RF{$6R^~^Ob7C-swMI2>XODR$Of1WLrOr_ZZ-mu^)lICd zwapu>=x0VNMQWsSBsq?~%6o#nYQ@)vZxCNL2S4wxU>T(ef2T}+o_b`%tnUgIDZs5C zR5f1y1Q^bM<&~r4m&?o7fPVtOKXht+6gVt*{FtrbR&b0G^86NaF__z_* zlP)Sl+J{))Dh0qf04Re4o1%3-rS4l})mR^#E!nyZzO_{oM>uCS?ZpP*p%O3s-w}d&}-vyiDv`Ux@ki-lx8npJ~7CUwW+FzYdIz^NrIlxRQ=N+ zj3lwq+L`FE4SC>L;ilJC6ai$;wT(J*t$WC{(T%-nVnNr6+(O252*6+05oI8ym+VFD ze*)$*T?^==A!fSUT$72}hNJ9BG<~&rEWXN_p-)4b?QK(4R0;c%%ni1gCd(k75DtBn zNT`M9IDgjs3CQ*NVgksK0ACyc#McfK?0=X75Kts0FhpcRA{0SlMsQ|%UA_N&{(p@~ z@q>H<{?+Wv=9^4ZywT6JBvjU$8UH`^l;~PA3V={e=SBjoj)8` z1jHkWHH?4z~N)JHF&d^+>(I&d2L%Km9;hKBkC?uMiJ*ll9?on8-!0HOk6pg zm+u#r${(LqLmZkEWz=A}dx4ZqN;+uHZCE?YwaSIU$XUn?E^6xnIY~v3pUESk+V!VF)j5BDGY7) z(Hm&y80{>+g2fpv6JgRAn&n=ZzbUo4sznkDV@3>Xal3M#{j1LCB0d1S;yzhoUkkC5 z)~4WY!Knw&OsA;2_D|di?!6ErU;Wg1K%?|uee=F(_AfQjt4H3|O)c;ClRD-XHUqnV z_8k4|EN)hR$duzp2B)oRd52%1i_ZgNX6R$4WJwLDGt<(nhlNBG)yPn%_e5Tf3wEU- z-3>&ORHLY935dHF(!$$>De#isG9*sd?hi|n%JA!7xKO4v=#n?by!219iEasqZnK&0 zh)zlN`Gix)tv>=CyOijdM5p&g_z5bpTiYBH(9=JueKS?35o`PQE7~U65 z7IJp0BO#kX+P}uT<`E|-h^Ea3PrqJZKJ;XcuS1$GHwiFECXUHGi0)+YCtE?8h#k4- zk)n8rHOxxX5AadMfbk>BEhBlnxQ@pf305O5D2-+v>=(qhj8-@{T@p&nV2akpG#|z0 zS$)BM-nWVW31t{5VU9hw0%?wPIK2NtTP50EmsScb6_lJ_`wPgqsFFx!JA{X8^d%fo z9hkwdOLL>IyR!*qc?n0g48~F0oswewm>-wDy)$%3hV+bywo<*0QfftgaHMYR_4JX(@)pJM0V)gH#jqN&_WK^m$X!4=WD8}8~R`;-4a?*igFD|xSUW;EU+`v z;2&`Mrgk{|n=B#pAq`)yvb>^}q>NQ>M;6_?cw40$hYFR+tm$>+v=*Aq|*6^EmNpN81b>*b% zyey+vM)D9c3V)zeqjVI0si!Z`(D&yNJ=9v(7L4YxC`s9oo=2vU5Ir<0Q!z=`g;SdL zht_PUR$|z#;LWG(3_&pIG3k+VetchCqpe#hO-uc^C5ix%#NAaS-zU zJ-W8F|3jUAum7cyu7vkZH{n2Uo$Oyp4Bza5N>8$=BcaOJ= z!5w8C3^wxMQW&nVJH~tIlIVeJXlKk#@@b~q##1Q{D{`_U_5scdhoJzae*`5!21p|E zN1~tqhNiU58lkEJ-A^r<)=P;(94N6oRLMt~5%Gx7?Bp-PFE;Bv1ioWFupY0tjD_Z$ z9w614(&Ns|4a{<;9wZ;@an*FwhdC?=MVvHYDX)pMop#EHZm$+hj?JEsnBid6)5STO zstXC9cW;lqAI&r}G?6XS^Bs?V&GkO>>Ye4M#45U99< ztWa9WF?aDTW^iX8* zLZ&n?Bz>zeoP4BOePjpl!*tEQ^!Q*sKAgM(C;L=~qTC)k$J97e&%;e*F3QUEk%<0X z>vX7Aj{BfVW)S8QUzNTsP*oEWI&)y)0W0}VW3s1a_eIemzmPtrzpoK1ejlSfWu1r6 zzBM&H@>z6rWvi@C!Tt50r9Y=6f#w3bT=w#R@^_goi;QvomrE;|vK7--Mo&>W3T>~|iFhL!3L@4imR=;zQ!?hs%K!Gmi2>8oTXo4@q~+3qa`{=wT^+gS2#<0)GTYKeq z;IfYH;e*dJNOMxQLu2Gd-Kt20|EA&TN6!roTlO4h;A`KWukSf2Cy|U7Hta8NIf%sf zWX+vr+)|UzOH7hny3j z$qjOgPyGaRY?xfvx%7@~!BBY8X8zw!>FN4_ENNLuF-@Iw(Pi?jhVwVccVQox0^gCi^YVKxoohbASDvPp&P+qMBaPioG{&$RZK&=QnjLXOhsW{Sz*o* z#Zr(r-Ot?9Cdl+`G%0s7lAQO+>^qlL?MJnj#m2HcVBHZ`^ZL23zbKlqk>KyCD?_`B z+-3*QY?{if&K`$rUm|7JTa4`%Tw=8EhbXPab-iJH#3#YXwZse_K_TqM#!+YK%KhuQ zs2}CZ+Le2-gP+<}1LiFF;Ya6n`Ogm*gsCLewjnY)oI$sQoBZODg}2h36Bv!3@8wG( zj3QIom^__ zpB2_cecj(SArlm2xWL6`=eJFD=$`57t&~2fv~KJ~Ug`s#GUk*yzgi%Oau~q1{jck; zZ>NwNvRLm{MSOGdiSFdC&gRQOJI1pEPx#KAX?GBAdIS!U9+;BDpA;f zMArNOKH#hKk@#VFI3=Zd6BD25PpzpYu&9ku#H}Qoc@R9|3mEn}7~hKPv&l%t^9dub zTND*Ui)kMU_K$^=kfCCB!m`|df7K+JETu2aCWv*j?2LqThTinYNn0ZeTjDeqmAUYT zumi#*xYT4@ePkP+8tn8T*Vzfy;M9c*^wi-GVghEka3;CEe{Qgw#O;8=&gmP69p`o> zT*u)p5aKv3B;jFbkd(qa!;SNEes}y)P&&u5@k4)hiA9W*0@LPVLhh2QDA}QJ65-1Y zgAv=EjF`$HK0#R=txW?~>=}}A1LN_^JbtKrt?r|-Us(lBr5SD3de41#ReQsk)_ts= zdg5;6HTKy5YLQJAU$b9aQ1tBEvP;k#?6#{ZwQLxo^sW7nVGQc2*Sa1tcoK=IHRK2c z*4xR9#teiZD61gYNYa}znaE4!t0THZr{E>GepOsn;?17t5p>TxQdq3h&hNYt>xZkO zD=XJ*2;UZPCQQ0lCTnSwmre2D>_KL({EH@`*^%~XL=(sgAedVg`iM${bP>0?1CuBf z9#m&qa(|Eyf?h7wXb+3D>tRB6w2Ujt$j{i)rm(V*BH$5qM%Fktq*5jMCwJ}}kgQCv zhhG;u>*J7nB_p}eM=Ro8>O5?ZyESi94 zO_Ci$bisccT5a1lWKtGRQF|3QAxk#vf4qw_4R~TVZCpXiw_ACY+Uv!{ka>_FSR2@} zyly*nkAh4JyO&?RrI?fohMiNJL1bR9pRADMOp4&qDiroZ8w}-<{K%f|F`I#k+f6;i z?D^?M`W02hC5pwlB5&m%&FS6h<8hEt+51DSSai8dVzO{F7}#-eb$0j&E`I#$8h!f_ zBRU~e%Iv>6)R47(J#rzKl5;WCQtwK?S+8Wz>#a*YT5pw_`tmxCu@Y-mbX^$VPW$gl z(aD_5Q1%2P^|*Fr>+mzHiN(s=E4lMAgHyGH(dK|fxgC6=k;9{A12-0S7qmJ%VsGZ0 zPwOJB2XTi51#Y@aw=^Q`yNB2)(qeoBx6lS;8m?>aW$>Pfyo~htd-St@^JGew2%mso zw8+p>5|?_ zJn1nQOsmaC$De6TKiG7R?2!ACc%$LmL$S%m3EdqOO;;z)3iilOHfBb8TGQ!@WdEhiuR@OOr+yTxpAJBWXUut86CByTI`;dwMdI)r$bwVAr?fbV=FJz0INK{= z-EnFFxEwBhb7aTeLZiG=+t-tJC^wP4Bc@Z_K{~(qOiTP^Bt`8@ zt|RBujUz|l0?qXI<}VAvgMWNgN$FHhzPaoYy8OVfPLqaC{9gdpBSe?ll|bodCgRh- zP{O5HLGlgUq^QKP>`6T>#iFZ1)wT+%$<*_R5>Y|{{}+E2_?*bfRW7x{|0OiUIz>1% zK{5wYqutq1)d}HNa$ses0pgk#1#*Oik$yU1M9Wm1=jB4rDh6V(Gc7kTUj{bipd!~i zS*`ilw*L7Zm1I=&@}LOUrO6(+klVb&*~~NoM`Bp;N-k_xR(+;oPDqgGTCYK;T-RO` z@)mqZkSWEXKsC;jn1}@TZ1E3fJH{17JAsV^3(}5aIN`%eeCH;R@W--L-|L<>n}|Kj zBS{D2im}Of{Y$=Mz!tZfWQt`DD{@f>7&4~p2WXj8B-DOoERTh8hbmSsS;MTU^dYgD z8GRERdRqt4Be5E?2}~1e5?Y{#P?f>K+h2hvh{Giimn*F3@4CN=*bcoowob!>h9*4y zU5n3x0Yv?T1`zWT0JKJrv|ngdD+b?nFLSsJ3#M2ov-(QMl&SY2GDYRr09`{+Q>07w zL<+9{SFHvqhUOZw6o4Vg2}n-%RBOFRb&nlS!>SCAOZ=l6WYP+NiZi=?qJ$Ma+W?!2 zCY3G9zWj@u*M`#aiMP_>Rw74-H~J(q)btchkP4d$SJW7P(mjM_O{_w%W@55!LF6aE zd1rSiZkJ3|5^9f4w?Mtc3~xT-un4*aMTA(RPAYzPY!u|?(Su5ew7D$BGDM;7V+k4| zXGJZcXz~He+VC+1{%pmh_fhS-Gs%-dFB@k0>jLXP7e?7It2VulX`M~gmJC+qYvzx= zkE{Q*{zvfNM~t$erA@(_w*O*A0ZW5Vbq;h6EH|uVwNS5iWGa#qfvN1J=xedcq;$z4 zZZN0a*|sT%2rWLV)E*!-YcOjtLAF(B$&de>+R=pZv!#bHt|GYjy4PKp|0P|6S%scX z{{$@SeDPYeh+pIYD9Asq0OH?Y68@0|NJLD6h(Glh?XLY|b8Bx11$t(7{*Ree{6t?f zt5DHTJ@4N7LVswGjh*bnH^`Ba(K)u-AAAjBB<{wtiwb^A;FmkL7F16_5kM<3&BA@r zPNvuFx)2~K!n8ABywo~A$K<1C4O7!Anb<_#dmmdZD}gzH9ES4hMUK1C81Kul*Oo)! z5Q7RKq@p*mk+VUnk?o!Qwj_eBRIE-WfBi88D3AM%J1#@B%kPkZ-mm;Z-zcN$xrY1a*Z}l6G0OyyHFRB@;|>Fp!(5&l+AO z1y$vd0=pWMDct#v(HNO~qZU93k#|GieR&*d>zpFCskZO(lr*TNaM_zo-v~njVLF%- z`^%t5q07ReP>R63DM(8yXzS4zcd;6!m(;-aI^kEZ%}Z%dNry`cXQN58Jnh2vNB?4_ zb`>btvI8g?NewSFUQ>iYNe^oMP}gJX)p>VIXHKMdn3z^&NIfzF#Zdqf*X$pxN;{4y zrV#fqdkNDgf#48Apz2Mi_x=r)`gTHr98c~*WFzUdJ5@gm9Yyl6$@L_k~T>mZut<|lwEqT7GW0x^dK$D*O8gs1+O%K^URN~N+y z&^S1)NN5wp!ny@{n7lMZbd&iz%VXFE9u2AS(JKZm(R-?-0sIHw^c3_T$ey3?oYH(P zzwMilO} zDV9~!m*-V_7bR63DG!nJ zxT*d6|KsE`Wa+yFf+z5!SrpCg0?e0$=$eRCP4N=|={kVdHAV1ZmqEI1@gjE<==TYT z4BgrZ{RB)JeNAfb`#prPt$X#N_niDStvgtvCFunQxUC|9=17q8M~|_^_Bb^VqT^J( zYXte>*<^I|6VSv1qZ~Q<{cZxyMkML)k~LjRePW5=kvsvzw)yb(?E3(;Sb%X>4Lf z(b*SVCMr|UhTt}-lUuD93}@6aY(WOJPkw^zxZ znFPP77FCujq19;K{wQA}mSX7QSl3rpN8%KX6`=!2CsU#)~f7Im~_$3y3{M}>~&vi{yP&22Z z^;n(1WUJ%d@gS5J&MhHhu)RDd>Gp|1y++g5RfAj(6<+(40RB5);>y{))Yc{->%iZ zngrP>A@qfoK85(ykjLYcW8OJ*kIvtos`HoM5fmv8Sk@On{%(LjtT`Sdp~1Z(9ks}m zE=3%h{u3Opzt=$oSL4TZJe9mRO{nw<$m5YtKu^~M=k!-2wrO=xuuPn8v=Dsj&RM=M z@hSgxNW-U?Ija2Iy|>{Oo|9V7qpPQ+xN0aAIH?J*8o_$P^|*r!r(038MYrT+7aq>K zG4m&46jg>`;dyEfT#=dVh6OT$Lee0?`YLCM_uP*_-zp)yG$M&Xi8l;}v&lUd*U51+`mZXsMa9Q=*V^2;7sKrAhV} z;F4TPoFm-=!Dx?=IE~Os5kLiO?ouJXtoy`-`dHh8KsaA!FY@i(kA`r*;q9VTEqWl| zp_SMc7~s1)UiBrqi#!EWpIXd@_c_dF_bs)iULO!_(OA(+##? zm71`-5JDZ;*TW&%j4n{I-Aac=D5*6&4n`P!$78{3Ks8%wRP27s zm`j!^vAqBTFK`JuvI=`H51~J@OCaxM%CW21d(Eq4+=aWYR>|MnOi(C@ZO{4To)3${yHUWML) z+K?yGnIj;~^be?tQoiKR1Zgn~M2rGNbHf%E!m^RgAN$sztQm*Ik>bp)`N%Vxw+9zt z{o8KY1kJhYo+!hS5TUi-T*7y$!0~_D8-tJI>pkv|hCgj0EmRdTomgTT`xFIpz+=wu zm>WuA**z2N-}~^s7_l^Y^)GV>2u6l5*-@{kN+`Xn{q|<*OyC}YqPI%R3fFtgxMf`2 z)`z(wsJ7Og5y42PZW({CKYF}p_dB?t0rPMqu6!}l#h66v`zeoFiCnr_kDEK_+dqJJ zz|x^y)Ts1J8tJ>Jm^~BYcVgJlNMU>FxM-w`)|OnfH|f6Vx1*C|-gtwW{@C-qruvpOBso#6ktOOhl?^4J0t>?ff zS0!pD^wM=d=^M!q=L)g97^|dudO0R{Bw&gFknEr#wQ9YPQ5LhAGnpaiza?HeLb7m_ z#F(VmVeoPI0s3#hjympR*9bLQ%b+ZYEa<#IC z8X#~wH&e*7Zs1{3Uw}VfIMKxRlAT!9Ltv+kO6WET08Y_xDOZE*wtF`|R&1I-70!~Q)r4^b3OqGoX^13~+fzXqY*BJ2_TNIPpReualcvti& z2_(;hc-LoypZ5K$IYUl-SI|2|jT&jZV6bGyaLam8gKFt3jPd2aL!tF#!VgKS45#%W zEtCHGVPaaFjWd*c%Omy2U5PRB!c&S70qJJflSMb0nCv2~VUVp<&4HjN1p_Bx<|76u z4Hm9>BhEZ7#hs{ftki>=!+VJzfH(JmowbD+=~Ze%cb9Mu`5ZMPtnZh5806WPpvQs z-bpB3m?v=LK&PY->_Z*K;F4f`9MDtCw}Y`y0ACQLS@k!rE&a%uMZmd2@pxbSH!5D7cWtS3rMUsd&CMmL_Tg7Qw=04CP%wREwrb5 zN$9j*!*<0jLWl{5SM12s*9lkMtO7rpT4)}M*Ej-D6HOSwxbEu#dMWd5?v81#2(*<;N6y0Ao{xGB;F0^*qv!5LeS znSyD~`KO}W%(-;qQ>rrWu$B@gb;yJ}0nHGwj44U*av%HodsfaxLDP&03DZW&AtO(= z#rTb*R*iSGL4R`070U)VPXh5ibs>NUaYf8QkcEJ8vC-u6-^=AEQHYS1`19|*A{&Jw zdgNjA+7a;IruD5{Db5faID5>M--@+kQeYEzD1XXhob~IHgM|&50*^~&b};qp$WM8A zf65^xJcX+gNETBG^z&d={5_lIhV<$0kj7w{GvsTJhp37H^92Be(66IrE60JI0x2$Zj-yX1ZrmVK!jmq}60 zs7}X2%)etsaXpHO6iIeoLvo*OmLywxmB}|xRA7sVN&}2)`=<$SQQ$=DLg@a6_w(B1 zp_tvhxdYa0{zl@;e!qb?1W~a!b}pneay+EbDmh@Bn;eO+jOmtW42o4TIVZ>8!Hl>c zT7lQwfpqo#8hOb}nDvtOen->ZyT3nS3i|2w?3y{pq)?ig$9GMygs!rSlFcK>9F&ld zJiD7Q6|y&774!*+4uwP?A6?F%O;Ls==%{!Jd@?lRNs>lV3VI3+dD}^1uK>Iwbpvf& zcZl@!x#wCFnJ~wqZ;RXQ$J@nP8~C%uZ9MJ=h6IH-dE{DqQV~bB0%gfX=Fno08L5qt z@GmK44+tVainMR5q3jnNsIBgna*sTCi}Fo6Je+`R{0*@&=v$!Jl=0*Bgr$$5TmkH( zATccv&yeYpeoSy&O0*DQ-$N-qdfsQZkC0o)T=X3cZ>REMY+2Pzy>_aFzjBKo$E^dQ zb@$LV_MwHt^x}S@C~z}x{o_wLkd_8a9mAqC4nA3(2uPm*#C4e**>M6Cf&2@$s7w*T z$=+X(n%ZPF52ISl$$?747EIC@sH~ze;g_d{4S zEzkEpTrcim#F``7dxhedmCTF@1`o&U$)J=*B2Ek};N4z%4aKewU$wM^dLyY^ud9Mz z)9h37IDp5LRVZiqCO$%NgxvyP*qyNOAmh@eC8Fi@6V);ExTPK9f0gx>QE@auySU5Z z4vQ@APH=Zw9D>6VAZU;PK^ItHkpzN=0E=62m*Bn$5Flu92<{dV^4@*l@7(+SxO3)5 zot~aK-BZ=m-BnLLqbeqn+bPV^yV(lfcC}`qgN8+ZnEf+$m`{6NTzoe5ZF@}9&q}Y= zrJE;D1baS{B3+Tpe}@^`{W_Ug_pN*DwWO7TqL1osaJ(4KFU}>Upu~^XS^At#$_(SM zCh+G(_Q40J9}Th&v+OL{{|Ef}$_Hd`?uXi~iFs|tN(ZZC9)I5(MbKlY5B>~hp^3rA zXPZchW*!qBcU3=Co|?j%3S!hH8t!1% zaZWR!;Felxz@0lQY}IwYlB3a8WhQhj1#J_Bl`SgTsO#n-F>AZ*`}&7$kis%kp0piEh0w< z*`ZBy_e!89c7V4yC;md7{9#|gf2Pf$aB)M7C#5>MIrME~35 zfwum!xXu`JCQ^Im>5n(-0Xeynct9KT3|J<&0B^RDt}J}%2lYBz55Wm=>!qbN!v_q; zi|r&q73I|*+^Ze*aVcJ6SJKYdn8VS5Q1P3Idf|W2#&*=6FDzle3$7(j&!(ea3|3^{ z8z#P(Y##kRO>-{C(sL1U9T%Q^&2!0b@#1Wn@SL)dWD!kq;+u%NAI1w53a<}^PWq2h z1npl$6X1R${Q5+Ahe!e?+;8~qFg6SkqlACYOT3|yS^NKFb<>9jlj*1bDPOBjA#Z2+ zPgY_a(1{)Xul}0f6r=i>u=9{m_92iV8wq z)K~Y~#CK`Bp6JE;C4XGr%HQqG6Mr9Yi(rtLNx7eJG|o9Qs_w;exH{yx^Bvy+EAmfyTS{62lt z(4eggQrq(seuw(P7tWVh&8-X_*sefV9@@t(jb zBjqNAPx$>E<0T?an}QgKYQ%fH@Cx0Rg(sq@m(tAHM4eoqI9FR3epg^42yqVMK)yN; zzyJOBWiY2$M7TAoes&R-EN8GXY_CuL{?fe{nf~RaX!YChyFx5067&ygIG*<5_iZ2? zG;XB2frC)^T^_b9yzQ4{KDtl%-SvSb_E`SF{Y78+U8aq<-{n0aj)7tL-7-0;Et+wi zPbKR4-*rfY=30$CSW)-hyo>$GBesDzit2EhlBt`|9>^S zD*vCJQAz9nujv01P~LAgGL$u;(8MT6=l_EgqrBe$pt3LuB~=i;vW<9rZK=+5W0?JQ*Pm?TNd+K4mVwFe1W$_uKGmw z=m8?#Yah_bOT7f}>uWV~MKjpxn!u^pNsaA1fFq<|T6Dt$MRwWfM^^=nkR^2}1ENL# z^$=%b3J1q<`C$+(R3)ns&45~?Sw{R|aJB)q;JF6C@oIZ8IwTW!i|>SF1x^S*CJQO$ zH%*Sd`8rDF3l|*p>fiUYRY499h5|=FrNrA}>6y5&o>Z2KvfL-cHqL9=6l*gKd%{u> zBK|`a(Lyi5AN-(fZ0wjtz_cI+^6s1GAbPc2de;(iYU0P^;hfoG1aHEhQK&ode#Tr? z3 zZPy?3PNJ@@WqPT;>2p;pqIp%(vF&~iIG)FTP>5i=(7uUKWya@(y&Y75#T9P;3zX;h zU#tGVNO|=Cv1*i23o98R8ZZ8{>wi(oD7zN>HHCEfJ5h|7G}D2{|DhMHH089(q$2&r ze}_$MxS7qdi>UhJ0H6H#7)z(9rHF$#l|>Ym+=xeHhGmAOAJ?cD8SYvR#U)67{q|W$ z5}Xb@4!Dz~e)m%K)#(0aGb$g{P+;@wbUrHj+HL6*hG4c^EVd}J7>LB@>`yu5#VF+^tb*YuQ z_;Ypx?#hcM2VGj`mt=i*1=92}WS}iLRwjgT95|lpG+|+{_wq&^wp48{g#*i@i$Iw0QrG;oIHf?_2*kq4Dd{2fV?Lt!* z9rg^tcOFSt3aszZKRMPAm26e7RunTQgtD-H7=}7gCqJm%NhG!mm$>~9Ub$U1{rC@B zTBEr=Za*i|#ycm&b>-$pQc_rXlqw@p6jSvJM&#f0xv3Xb=C)Wyb&aXWWF?uolRjeaHY>D^!DK2~DJ)2bm@Dn7 zig@-XO7K04`&}XGbQBPUFRBX^)=}(Ag-ZP5+6_ogwaxRaIw!DB`qTL>vJ^E$`2Guf zVxVK-{7Y^2zhO_*0Eq?)qoSU(fD5I8t;_%AlTZYpqMnT07z>*jC-N>%y-Mw!iQ3vO zE;vwfTL`q62Wj2qC$xCT!?SPf9ungf8P!VHiaWEEov2iU#hfXuVbk%`7l^YENJ+w7 z%U~U#BEXg*(nHi_49vLZ5&EOVuadT#G?&0>n=sA*BMiMYyNzC&jMpJJx@Kb9qtT)C zm7(E>Z5g4(tM46A_o_O?hHIar_svJBKD%W7(&ppEC))IwJNyie029_xrPA87W!|cu3f` zJk{v8{;SR));Dtp{6=ky{YwgB|G0Ih-!oB5J64R;|MY*=cUPy zWzRRhk|HX03y*5?fthx*u~K<8YxJlEa7LVT+x=UyYD)ERL!I9$KwnV6AzR$|!{)`2eZ&AGI%BZV1WK~lK-%|!*AxmpGX$JDt>6s_JD{Sdqe9d@`# zFtfbZ*eQlGdcL%gcBD^^V`S`H)1zx&?h(Dw;k1U{w()YgWB2!{9LKXXDU5xS^tVqx zkC@9h*tPw6{Fj|bGDoi@Dh{x~YBAB7qR{9-^~zMfFcJ6)0yv-D=MF+6qR|Ydd9W7Y z-z1nN$-n%I*^zMRX8euzYiNFu7G?(njK;lUJvZi8{a*DNEjh=sHR0`WXCR;O2}8GL zTI5!9iH8vh9b-j5TkVDg0?g#C4 zbO!Rc>(?Fp@P0m;Kvh_zs;ICCqaGP~+L4!~Qgu4r_%49HKt$P2)7I)wfxYcz^a>YO zBpHpO+t-)=%&G6U5;upA&kCj?JyJYmS~}x>5nPwLmi{=0P_K1&v!?sIOwp_--$jkJ>CI5YqRVluH^P58Tj~IQogn!Ug zxYM+3u5as?sLnoVbYP4(?S0YzlYfXIf&zk zT1bd#{|uV%i&t4>K03Bf>}-~{Iv8I(aP@g>soenCfn-SE#(QftN!K0sL4Q?jGn3wa zsWKJzL=pG}i?uanbBz7vx^-G^WIJrI=N=1o9=Jo1oO@GQuZ5;BbV2ER z)5y+9>@T{YDbPF3U7Jq7Gb$V1Z2l}=*(`&7<7&qHoVPRCFOiCxS37SkmZxnZ_|%3G%rh`ZOW5=#d&YRHv@l*v#pLIiZK=+FVShj6^%YL zUOcJLK9bId{kT;$!U$(6)yaK&mQB@HVRS-zJiPuS^i8_tO+?xolba{D_8Ud5fXJF} zXg#XoL!oKk0s9^{cTy!iv?KsUIj=!E#u_zV)Lrb7f)C zFg}@9ZL+)t@J2zLYn!qHM5iej^>iT(d-&NrHiaW-l=6NIV!Xe@&6H~7*x}b{A z3l6Q}ULF8juC@%9mkcak@Z&?2k4-9kzMCqN@bY1jnZMN)Qpaqa|Ex^xMDR_s(x`4f z!RGBn983Jx6@e0CyiY96wv9bd(4jGEd4l^e2Og3!GhSX|UrNaw39GpKsF?J3oZ*>4 zI*?EklvI-3OBdy<*YZhsXPB%=g}>~3hyQNYa-4Xed1nWcaD#NUZI{qOHd&OkhD@eH zT6^LJpz*TFgMSXT|49ekkt9qwwCKQcwHvv#2>=OY?S(!|F@+R{6?&-JW6E3n65YB- zkNXB1!o#Xk8Pq)!joPN;Xxqs_n-mDm1q8Bul3UYRLMcMLTTL;;|x!q8L$&fjkNR5+JVl{SqLVQ zkJSS{MTZ(O2u$8Ih9WVaq`dk28V5$t&%6-2pQs*R_>9U3v>G$ zh?%LpQ5kr(7PYsBYP2h=(Y%b2?Skt|0|r`VIWflN%DgE#xfrX_iRl)6E}zW>%_NfT z{YnBizgjtzRygty#M2H$u*L)4$R^ZL28N{gWh;`cDGo{;D3ZRT`ArOr60O6``>h-F z^KW57E3v?v%Z^RbYz?kc-rFtZoaxhNhAcySnVC~{?DeNH@~-G2Ow=vUn+N#kHo(2@ zqBsJ~uy+eOm*YPx=S9yoPM;0stVP8NVm~3t{_?Xs%;uk2!M!UBm7gWkyM-G$1V)an zbFCIK1Bp>S4v+!rpZ1477(CMkZ5{=pI1jRSk~h)JLg znU~6M8`e8pCLJy1;LHT?@RAGIXQBCqWaF$*1Xrl&yR5jIH7pPIkB=d772ZqT2~vMJ zl^5!O#b>!M=hfEy$fXXjW3c$hHZHXk1HHDq5j#g{2d}Fl(}CpAX?pr=%|Cf$gfQ=7 zkm5q(y>`Y#>$?XV8Iwicg7`P(nIZmGIZeE#M39xHg;FIW^to5>kv-Onj2=`W$(8i4 zjkg{??AaA5>BaUX3l9rizG42$PLM)ith_!1cQN=QU1|hGh=m+cNGjzQ_BUaSXT23> zrPAy)ey1Gm>w9&RO<2ta4$1FI23PLTs7~@wiAy_RN)Uz%rwYMHDNb18$4S4LS%eq0 z^`^=~&rKNEs+=-%fTa6i!mmB*0j;iA+mf-OeC~dYQBrHYl6|84;6Prvs|~Kq1VHFa zUkTs|_oad1n~c=5t|4ckgJf+f4SP9$!rG?jyNvg~@Fxx2{5Y*xaDced@$rxahS%;q z`hEPZWlHUk5qc# z==+ceP;r5g0OXK{XNwjp(cbaPa^QvUkxBLbe!_~9k)kiK#k+&`$27;x#UCyf27l8* zq$QV4AoTjX6qame-N7-KLglGUwc|XMwhOT|MMIfi4UotbmtrLSevc#GND{$jyn93A z)95M7t&B%#&5fty+^ z@TL@Vx6V0(G)mmm$WX2w79L|a<=;$}~je^7tjG>mSC{bs8ZrjBqRs>wxgM6dvN*#qIrT$tD; z!_DH(wf&O`w9K~s%$4lS4Ga)>|A}%LXcBM6r*w!Y=quUJQG`_t*vc(z`bMGYoJG5y zZvdn9jc?&bj|h{WszNPWwM($3Uia$w;gxzBu!%@eH9j$O&;kZZY8c9{^=(;_O;L}f z(dW-)Vc<{eA@AftKlV$ZJhDDIK>=fTn`!1>9Xc$0es@^z0Ag{WrYzlQUI79nFQtcs z_hKEKR`RlpM+Yl~K$?dhm6_lT?6R3D3Gzi-mHFUp4;j|$rz-vFbakSfa!%$JUTV!``liEc)yjw`;1G&c89d2{UF5ma6_1+zV)9;jP+{2RU7bm=Suy4` z$-;z^e96*zcSDoG%eXmczD*aC<|m~~;zHjoEunJjpr>rT5ZlAMA$zXI^pdx*!qCcE z?j{F$)*qz=T5SA=)nuVJ$%lD$MRO~R39(ovyTKVi2s1BFFB|tDm$161gGSC5hmpD> zppe?!Zj@?E0Gl@Z-<-D+!d`RZ> zZU4q-5%2GdnFF+P2}{c&lFquVFbYOCcT8Q!N?iGi9%0ge3KNKwI6S&LR0z0Vi!4~F zd4eebb^uC46WrvvKq%27B~yRsyceU*=P2qD6Mb zF=T4OLx7>!O4T5S)#v1sjHeRdA9Q`2K$9>cdB2k7Y@|a%W7>jGx)6eYIa2j-{QUAy4Wk;KEkXuJVX}i5qUG)mSLf-15V&ia6;jjl0`a z;<_bFkVNa`VuWln5zITbb;VCAP|S!=uwT4dR*vklb6Gw@w7OT4Ah>FI>p46~Mnngb z`f7(^ZQOEGLk0#S=L9s~tA$>)8kv&|c`3@NoFh4K0RI||gq+F|fQNfDXvi1Hu{b(l zEqa-=6H@FNrUbJ0U-Y%X*~@hv1u-!vT&|UGS;Z$~VJ4a+nuv>S+_bi@Fe*Md5r4Cs z{El1r7=4R=$BO)l0s?q4+D^Abu&`sKbZ8fuonzH9s+ESV_U|sRwD+x1BEww7=QALC zWZ;)oPTv!hxjnAYQuZcS_Lo(;m)dS&dh#Rj?3*n zS(*1STYCmOHj;2cK6iyANkm;^P@@lvHce&Rg6sj@jZN)qmXEr+YD3bHgLC&zo6GWv z_9N=2+|-_U(PF7v@`(y)BLT+yaFqI4MCS1iu~o=D1njnpx*+ih(iiuX%Rq)%R!qjR zn?N+43Yer)n23%;AWtqRxy8tIV}0ELZ1O!adBoRLF$#xSlVga_OH+BT)9PMf>LDza z0P(TZv=-tdZ~a!llz}DB_Hmf9AdY;HyS7VzILQ?}?|%Rw?wIqcOx1=4i@m>-j-tK* z%+yrn2s?~A3aRN0IjB1Ufmh@baZP*K#9PG}h`?Ve?w8l}N^?R}aGflubJ^Prl{K++ zB3=ayja|&|*zJ2I$N0XL(#eGr441bc>ey?sp<};dg~sc+_EToqX8ooPAqr+|g9;#{ zr%n?q1Vaf@h0e*p^V$j9K~Fy!Sna>4-qvii_Xvx@#Vb<)h@Vb&5Mo;lSHoR$vvLd6 zSV;})flXZITM_(3>@snd)`MMqQbw7TLO+j%)&$pO3YvzL9{Lid*?c|9+arv6}YC;902^8yR^b6FfXj!C!Fo`#>ga~yGR(7(qrFJBq$q{w0q|S zc3hoS3~oum%XK8>R z-1J*GCYYf@5FogCe49GakGX#og4GQvhQc)Uwhs$hF>&HTbEV&C>-~eKav;YyFSZ9^ zgB9O{lQeVPuK6k&N$}AJw5u+gxo10AzI!wIVM{J|_VDJbG@I!ZxnCR~3i%8G#r-p_ z0`!LQ)zPEszB+mNnn`$o4D3>qmQ`A(jjt9&?~-iU=&$UsExJFqibgou#=VVGgkjPv zKWePo;IO8>dSjShK$ zuK&>;T1ftWN}omRA*5tyb^n2=zas{7nbagsZ3<($t7=#_CBjDdMTp40NyR6_)k@M~ z>;6MpLw6Q8HA}z#Rl`gu3%GnL0jK17C`?HUnWp2ZpAA z5a7s>Y0jF^*{*3Eh}yG-JGp8eN#&<3)t3(55=$=iV%@`t5=m#*EGzfQ@mDE~xs~{w zg|x0Y>`RG4;{~nr1a*)ZOs-_B+(yFj2^kw;GZpwv<**s+?KaMdc~oRolFF84fhv9x zj$0N9YL3HG>&-t#i=9|&O34MX2#D~rXnXKc_Qax@e30-q7W-}PV6Sa_EV3ch5xd;Q zXfsF024P>pV3M6S6D>_BeOBG2l*8;DbqHkkSiVFxiii}{&^Ri*e%?9p9T z{wkd@Ls2uq9X4pFfUVKW!f54U-W^40YJE5o{HSW!@5OBC8RXU02R6b_I+ADbid$l1 zm?{LQLe7NXj>aV_=S}XH&5vi-AE)SY^j>WtB&E}~txIgu>QwXwJrud-eTPnSJaZRR zigyUs5NR~7a{@n-f9n^2*%xl_)Wz9s?Q(9FFdW(y-Rk^-wTW#G%5*(aqU~*yOS9vH z{?4dqc*CfsvtVP2jg_slslZ$&xTfpofY$mri0xN{blNBl`dP_xOT+Dh0@D!lwCCzH zSvRdJX{`jPzo9}Txxw=joegH`8+qAvU(XVUd3Nfk<_Nr7NI??J_fHfv8r{a;UNY=p zByqgNJ|>*pcjo_G$xJ&H2^c=Dz9i42`|&JS!#-yyp+uwVAuOsk@UW@})+2(knfqN6 z?Dslvh?%cD*u;gZJIfbx3|%y#J=Q3o`y4!-l!Cmhf!Jf7S8AD9zmeeAqBI4_Io&CW zuAOxC4nsffTN8KHIX@`dHt3~#f2EZmC=09i3y>i{=P$rgwPzsCuX(i=XuLJ=Fy8@B zw#rcLcxt01{Qja%L)<~nk~U-RJKh3iIGvlA($WAiKZ(O2<}93RwL`o+6WdgE3J*MY z@e2$5dT16ZR)Xtxf%z^+r+o_70au2MhBL#^PVc5;2kvsR9Wj&70z<~HaXxY2;sR}P0=w;#ad+M8zf*bCag!7e$8cJCu zB5cv0-5!>(K#o0A<5LT6ogr(@q4OuM0?J0=WMl@dd@p)zjdgIgqZX!J zY58A9-R_&=;Z2z;|1$jqDrM|Rs~&UTLTyN7E1REV0OLx+7zPR{KW|ef1 zAbYlLMrB6bax~svfzcvUYl3WnDS6j+ZGRaY1Jx|&Oe!*}F(cyaAYGFgvo+Dtu~9+n zz7W}@0O-u#M<$JDI>npfKbuA3NZUcs5B!;yVV;9j@I2>6%nl%ZPw~jQ&A`5;s{bM$ zRj?qi=ghi(SqPDv`$o>6yf%y2fMUbAGdEU12B{25bc4}V=!Kwur_+OSdBweyeQ!Qx zmv*DSPk`t7QB$d+q+i$5?YTEiE8#w6y9_^@ma1|L!LmA_;qPEAVGvnUxbu@BQ}+HA zM`KkcQeB^ftewwH5dBdEP_&A3hOhsMov!)l^&C~8#Y?Y^y<1Z(ZEooIj2wTle4~yt9 z9bJha&uer&bHb;x?b%+{kQ6mWsY(vHSyJK9K?fKB4$*VHA{qMq{04Ty zpU7fA3S=F*aYjgTeGHNX1nBV06u4}X?L)bCsZWjAJs}vcW?2#qGGnpKM4(0D9!`z= ze1KX6mT0dIK~LQ!9hQaeL=v1}=9$4LH!p) zYb<(>tg-Vsh^=P!N=QAoMPjJR{Cs=5ou2OY`!s*;U z-RI%Uk=Cn;i+HDmk<;mD>dNSlU02XB=@8i7dYy5DJmjkhv0KE|9RKyb!;DPo zAEZjJW6N{R+DHfSjTI}VT`6SQru)2^jBUQoc#O%uQ;~qFE*$F#e|l^=$>^IL3oZtS z(&-erjBc$v@iH@GOK^6S%&<8Q2_BvgdS({Ry%h7ibc`i1lbdOhQ>bdO=a()yO&#w5 z4xm$O^=3V#xkV=n-2k_IIcov#LZ!r__S7+z!mv7V{W~RzI5~=R-X^(};nzN!jk(d6 z*F4AAQEzgNam$6w|CR@ZbNqugZ1{)rOQbvc%p-o*->>wzCs9X65-u?a*u{L!3aOHu zdaSAW{IdTm|A`OgCt@h7R$;jhqZn2GZSN)a&(a4!c%i?6{pGYB?4!U!yQ$I__d^z! z8>xv~jMh3gV$UJ4wMQZZWwe@Y#QNRQqq<@#XTkHK;yVpioApo=~^ZtoXYU_J0> zw032Nc3!?J{yTGfV|q89rmbGMv##Dk6M=9jspa?*u1if=?W}MT?qrp5na;0PzCAH) zqRB3W^r%dsoeDGen{UO(!%}tH|Rrq{aMmUq>qSgd-px{#zMqi znd5E5ri5i!pD%&fp5j_}oc+<%t;4>wN?0RR91 literal 0 HcmV?d00001